You are on page 1of 49

(/) (/)

Journals (/about/journals) Topics (/topics)


buscarmenú
Information (/authors) Author Services

(/authors/english) Initiatives (/about/initiatives) About (/about)

Sign In / Sign Up (/user/login)

Submit (https://susy.mdpi.com/user/manuscripts/upload?journal=entropy)

Search for Articles:

Title / Keyword

Author / Affiliation / Email

Entropy

All Article Types

Search

Advanced Search

Revistas (/about/journals) / entropía (/journal/entropy) / Volumen 19 (/1099-4300/19) /


Número 9 (/1099-4300/19/9) / 10.3390/e19090437 /

(/journal/entropy)

Submit to this Journal


(https://susy.mdpi.com/user/manuscripts/upload?
form%5Bjournal_id%5D%3D5)

Review for this Journal


(https://susy.mdpi.com/volunteer/journals/review)

Propose a Special Issue


(/journalproposal/sendproposalspecialissue/entropy)

► Article Menu
Article (/)
Menu
Subscribe SciFeed (/1099-
4300/19/9/437/scifeed_display) buscarmenú
Recommended Articles

Related Info Link keyboard_arrow_down


More by Authors Links keyboard_arrow_down
Table of Contents keyboard_arrow_up
lens
Abstract
lens

lens
first_page
Introduction
ajustes
Hurst,Solicitar reimpresiones de artículos (/1099-
and abierto Artículo
Acceso (https://www.a
domain=www.mdp
a Altmetric

Una
Brief breve historia de la memoria larga: Hurst, Mandelbrot y
History
elofcamino hacia ARFIMA, 1951-1980 share
Share

porHydrology
Models
Timothy Graves (https://sciprofiles.com/profile/author/dW9sWmdNemlaVnR2R2lqWE5Fb0
Help
announcement
format_quote
lens
1,†Mandelbrot’s
,
Fractional
Robert Gramacia (https://sciprofiles.com/profile/author/MFFQTFo2cFpKWmUxeDUrazRvT
Cite

question_answer
1,‡Models
,
lens
Fractionally
nicolas watkins (https://sciprofiles.com/profile/author/TGNSSkVMV0hFOGp3RnZnK3BTd0
Differenced Discuss in
2,3,4,*  (mailto:please_login) (https://orcid.org/0000-0003-4484-6588)y SciProfiles
Models
(https://s
Christian Franzke (https://sciprofiles.com/profile/166387?utm_source=mdpi.com&utm_me
lens
Conclusions groups/p
5
lens
Acknowledgments utm_sour

thumb_up
lens
1 Author
Laboratorio de Estadística, Universidad de Cambridge, Cambridge CB3 0WB, Reino Unido
2
Contributions
Centro de Fusión, Espacio y Astrofísica, Universidad de Warwick, Coventry CV4 7AL, Reino
Endorse

textsms
lens
Conflicts
Unido
of
3 Centro para el Análisis de Series Temporales, Escuela de Economía y Ciencias Políticas
Interest Comment de

lens Londres, Londres WC2A 2AE, Reino Unido


Abbreviations
lens 4 Facultad de Ciencias, Tecnología, Ingeniería y Matemáticas, Open University, Milton Keynes
References
MK7 6AA, Reino Unido
5 Instituto Meteorológico, Centro para la Investigación y la Sostenibilidad del Sistema
Terrestre, Universidad de Hamburgo, 20146 Hamburgo, Alemania
* Autor a quien debe dirigirse la correspondencia.
† Dirección actual: Arup, Londres W1T 4BQ, Reino Unido.
(/)
‡ Dirección actual: Departamento de Estadística, Universidad Estatal y Politécnica de
Virginia, Blacksburg, VA 24061, EE. UU.
buscarmenú
Entropía 2017 , 19 (9), 437; https://doi.org/10.3390/e19090437
(https://doi.org/10.3390/e19090437)

Presentación original recibida: 24 de mayo de 2017 / Revisado: 3 de agosto de 2017 /


Aceptado: 18 de agosto de 2017 / Publicado: 23 de agosto de 2017

(Este artículo pertenece al Número Especial Sistemas Complejos, Dinámicas de No Equilibrio


y Autoorganización ( /journal/entropy/special_issues/non_equilibrium_dynamics ) )

abajo
Descargar Notas de versiones (/1099-4300/19/9/437/notes)

Abstracto

La memoria larga juega un papel importante en muchos campos al determinar el


comportamiento y la previsibilidad de los sistemas; por ejemplo, clima, hidrología, finanzas,
redes y secuenciación de ADN. En particular, es importante probar si un proceso muestra una
memoria larga, ya que eso afecta la precisión y confianza con la que se pueden predecir
eventos futuros sobre la base de una pequeña cantidad de datos históricos. Una fuerza
importante en el desarrollo y estudio de la memoria larga fue el fallecido Benoit B.
Mandelbrot. Aquí discutimos la motivación original del desarrollo de la memoria larga y la
influencia de Mandelbrot en este fascinante campo. También dilucidaremos los enfoques, a
veces contrastantes, sobre la memoria larga en diferentes comunidades científicas.
Palabras clave:dependencia a largo plazo (/search?q=long-range+dependence) ; Efecto
Hurst (/search?q=Hurst+effect) ; modelos fraccionariamente diferenciados (/search?
q=fractionally+differenced+models) ; Mandelbrot (/search?q=Mandelbrot)

1. Introducción

In many fields, there is strong evidence that a phenomenon called “long memory” plays a
significant role, with implications for forecast skill, low frequency variations, and trends. In a
stationary time series, the term “long memory”—sometimes “long range dependence” (LRD)
or “long term persistence”—implies that there is non-negligible dependence between the
present and all points in the past. To dispense quickly with some technicalities, we clarify
here that our presentation follows the usual convention in statistics [1,2] and define a
stationary finite variance process to have long memory when its two-sided autocorrelation
function (ACF) diverges:
(/)
lim ∑𝑁 𝜌(𝑘) → ∞
𝑁→∞ 𝑘=−𝑁 . This is equivalent to its power spectrum
having a pole at zero frequency [1,2]. In practice, this means the ACF and the power spectrum

buscarmen
both follow a power-law, because the underlying process does not have any characteristic
ú
decay timescale. This is in striking contrast to many standard (stationary) stochastic
processes where the effect of each data point decays so fast that it rapidly becomes
indistinguishable from noise. The study of long memory processes is important because they
exhibit nonintuitive properties where many familiar mathematical results fail to hold, and
because of the numerous datasets [1,2] where evidence for long memory has been found. In
this paper, we will give a historical account of three key aspects of long memory: (1) The
environmetric observations in the 1950s which first sparked interest: the anomalous growth
of range in hydrological time series, later known as the “Hurst” phenomenon; (2) After more
than a decade of controversy, the introduction by Mandelbrot of the first stationary model-
fractional Gaussian noise (FGN)—which could explain the Hurst phenomenon (this was in
itself controversial because it explicitly exhibited LRD, which he dubbed “the Joseph effect”);
and (3) The incorporation of LRD, via a fractional differencing parameter d, into the more
traditional ARMA (𝑝,𝑞)models, through Hosking and Granger’s ARFIMA (𝑝,𝑑,𝑞)
model.
The development of the concept of long memory, both as a physical notion and a formal
mathematical construction, should be of significant interest in the light of controversial
application areas like the study of bubbles and business cycles in financial markets [3], and
the quantification of climate trends [4]. Yet few articles about long memory cover the history
in much detail. Instead, most introduce the concept with passing reference to its historical
significance; even books on LRD tend to have only a brief history. Notable exceptions include
Montanari [5], the semi-autobiographical Mandelbrot and Hudson [6], and his posthumous
autobiography [7], as well as the reminiscence of his former student Murad Taqqu [8]. This
lack of historical context is important not just because a knowledge of the history is
intrinsically rewarding, but also because understanding the conceptual development of a
research field can help to avoid pitfalls in future. Here, we attempt to bridge the gap in a way
that is both entertaining and accessible to a wide statistical and scientific audience. We
assume no mathematical details beyond those given in an ordinary time series textbook (e.g.,
[9]), and any additional notation and concepts will be kept to a minimum.Our narrative is not
intended to replace excellent reviews such as those of Beran et al. [2], Samorodnitsky [10] and
Baillie [11] to which we refer readers seeking more detail or rigour.
The key questions that we seek to answer are “Who first considered long memory
processes in time series analysis, and why?” and “How did these early studies begin to evolve
into the modern-day subject?” For specificity, we clarify here that our interpretation of
“modern-day subject” comprises of the definitions of long memory given above, together with
the ARFIMA (𝑝,𝑑,𝑞) procesos definidos en la Sección 4 . Para obtener más detalles, consulte
cualquier texto de series temporales modernas (p. ej., [ 9 ]). Como veremos, esta evolución
tomó menos de tres décadas a mediados del siglo XX. Durante este período, hubo un
importante debate sobre las interpretaciones matemáticas, físicas y filosóficas de la
memoria larga. Lo que más nos interesa es la evolución de este concepto y el debate que lo
(/)
acompaña (que a menudo citaremos directamente). El tipo de memoria que nos ocupa aquí

buscarmen
era una idea conceptualmente nueva en la ciencia y bastante diferente, por ejemplo, de la
plasmada en las leyes del movimiento desarrolladas por Newton y Kepler. En lugar deúlos
procesos de Markov donde el estado actual de un sistema es suficiente para determinar su
futuro inmediato, el modelo de ruido gaussiano fraccionario requiere información sobre la
historia pasada completa del sistema.
Como resultará evidente, el fallecido Benoît B. Mandelbrot fue una figura clave en el
desarrollo de la memoria larga. Hoy en día, más famoso por acuñar el término y concepto
“fractal”, la producción de Mandelbrot abarcó una amplia variedad de temas, desde hidrología
hasta economía, así como matemáticas puras y aplicadas. Durante la década de 1960,
trabajó en la teoría de los procesos estocásticos que exhiben colas pesadas y memoria larga,
y fue el primero en distinguir entre estos efectos. Debido a la diversidad de las comunidades
en las que hizo contribuciones, a veces parece que el papel de Mandelbrot en el modelado
estadístico está subestimado (en contraste, digamos, dentro de las comunidades de física y
geociencia [ 12 , 13 ]). Ciertamente le pareció así:

De esos tres [es decir, economía, ingeniería, matemáticas], nada supera mi impacto en
las finanzas y las matemáticas. La física, que me temo fue la menos afectada,
recompensó mi trabajo de manera muy generosa.

[7]

Una parte importante de este artículo está dedicada a su trabajo. Sin embargo, no
pretendemos transmitir en ningún sentido su “propiedad” del concepto LRD y, de hecho, gran
parte del progreso moderno relacionado con la memoria larga en estadística ha adoptado un
enfoque (ARFIMA) con el que no estaba de acuerdo.
La motivación de Mandelbrot para desarrollar un interés en los procesos de memoria
larga surgió de un intrigante estudio en hidrología realizado por Harold Edwin Hurst [ 14 ].
Antes de proceder a discutir este importante trabajo, es necesario dar una breve historia de la
modelización hidrológica, las contribuciones de Hurst y las reacciones de otros autores en
esa área en la Sección 2 . Luego, en la Sección 3 , analizamos las reflexiones iniciales de
Mandelbrot, sus refinamientos posteriores y las reacciones de la comunidad hidrológica . En
la Sección 4 , analizamos el desarrollo en la década de 1980 de modelos fraccionariamente
diferenciados que culminaron a partir de esta secuencia de pensamiento. La sección 5 ofrece
nuestras conclusiones.

2. Hurst y una breve historia de los modelos hidrológicos

El agua es esencial para el florecimiento de la sociedad, ya que es necesaria para beber,


lavar, irrigar y alimentar la industria. Durante miles de años, desde los albores de las
comunidades agrícolas asentadas, los humanos han buscado métodos para regular el flujo
(/)
natural del agua. Intentaron controlar la aleatoriedad de la naturaleza construyendo embalses

buscarmen
para almacenar agua en épocas de abundancia, de modo que se pudiera sobrevivir en épocas
de escasez. Los factores combinados de la Revolución Industrial del siglo XIX, como ú el
rápido crecimiento de la población urbana, la necesidad de una producción agrícola en masa
y el aumento de las necesidades energéticas, llevaron a la necesidad de construir embalses a
gran escala formados por la represa de los valles fluviales. Al determinar la capacidad del
embalse, o equivalentemente la altura de la presa requerida, la solución natural es la “presa
ideal”:

Una “presa ideal” para un período de tiempo determinado es tal que] (a) el flujo de salida
es uniforme, (b) el embalse termina el período tan lleno como comenzó, (c) la presa
nunca se desborda, y (d) la capacidad es el más pequeño compatible con (a), (b) y (c).

[ 15 ]
El concepto de presa ideal obviamente existía mucho antes que Mandelbrot, sin
embargo la cita es una definición matemática sucinta. Naturalmente, esta clara descripción
matemática ignora complicaciones como márgenes de error, pérdidas debidas a la
evaporación, etc., pero el principio es claro. En realidad, como señaló Hurst [ 14 ]: “las
mayores pérdidas debidas al almacenamiento no se tienen en cuenta porque, a menos que
sean pequeñas, el sitio no es adecuado para el almacenamiento durante todo el año”.
Desde la perspectiva de un ingeniero civil, dados los parámetros de demanda (es decir,
flujo de salida requerido) y el horizonte temporal, ¿cómo se debe determinar la altura óptima
de la presa? Para responder a esta pregunta, claramente necesitamos un insumo, es decir,
los caudales de los ríos. No es difícil imaginar que para un conjunto dado de entradas sería,
en principio, posible resolver matemáticamente este problema. Rippl [ 16 ] consideró por
primera vez una solución convincente “cuya publicación puede... identificarse con el
comienzo de una teoría rigurosa de los yacimientos de almacenamiento” [ 17 ].
Despite solving the problem, Rippl’s method was clearly compromised by its requirement
to know, or at least assume, the future variability of the river flows. A common method was to
use the observed history at the site as a proxy; however, records were rarely as long as the
desired time horizon. Clearly a stochastic approach was required, involving a simulation of
the future using a stochastic process known to have similar statistical properties to the
observed past. This crucial breakthrough, heralding the birth of stochastic hydrology, was
made by Hazen [18] who used the simplest possible model; an iid Gaussian process.
In practice, just one sample path would be of little use so, in principle, many different
sample paths could be generated, all of which could be analysed using Rippl’s method to
produce a distribution of ‘ideal heights’. This idea of generating repeated samples was
pursued by Sudler [19], however the stochastic approach to reservoir design was not
generally accepted in the West until the work of Soviet engineers was discovered in the
1950s. The important works by Moran [20] and Lloyd [21] are jointly considered to be the
(/)
foundations of modern reservoir design, and helped establish this approach as best practice.
2.1. Hurst’s Paper
buscarmenú
Harold Edwin Hurst had spent a long career in Egypt (ultimately spanning 1906–1968)
eventually becoming Director-General of the Physical Department where he was responsible
for, amongst other things, the study of the hydrological properties of the Nile basin. For
thousands of years, the Nile had helped sustain civilisations in an otherwise barren desert, yet
its regular floods and irregular flows were a severe impediment to development. Early
attempts at controlling the flow by damming at Aswan were only partially successful. Hurst
and his department were tasked with devising a method of water control by taking an holistic
view of the Nile basin, from its sources in the African Great Lakes and Ethiopian plains, to the
grand delta on the Mediterranean.
In his studies of river flows, Hurst [14] used a method similar to Rippl’s in which he
analysed a particular statistic of the cumulative flows of rivers over time called the “adjusted
range”, R. Let {𝑋 }
𝑘 be a sequence of random variables, not necessarily independent, with
some non-degenerate distribution. We define the nth partial sum 𝑌𝑛 =: 𝑋1 + ⋯ + 𝑋𝑛 . Feller
[22] then defines the Adjusted Range, 𝑅(𝑛)
, as:

max {𝑌𝑘 − 𝑛𝑘 𝑌𝑛 } − 1≤𝑘≤𝑛


𝑅(𝑛) = 1≤𝑘≤𝑛 min {𝑌𝑘
Hurst referred to this as simply the ‘range’ which is now more commonly used for the
simpler statistic ∗ 𝑅 (𝑛) = max {𝑌 } − min
1≤𝑘≤𝑛 𝑘 {𝑌 }
1≤𝑘≤𝑛 𝑘 . Moreover, he normalised the
adjusted range by the sample standard deviation to obtain what is now called the Rescaled
Adjusted Range statistic, denoted 𝑅/𝑆(𝑛):

max 1≤𝑘≤𝑛 {𝑌𝑘 − 𝑛𝑘 𝑌𝑛 } − mi


𝑅/𝑆(𝑛) =
√−−−−−−−−−−−−
𝑛 ∑𝑘=1 (𝑋𝑘 −
1 𝑛

The attraction of using 𝑅/𝑆 is that, for a given time period of say n years, 𝑅/𝑆(𝑛)
is a
(dimensionless) proxy for the ideal dam height over that time period.
Hurst [14] then examined 690 different time series, covering 75 different geophysical
phenomena spanning such varied quantities as river levels, rainfall, temperature, atmospheric
pressure, tree rings, mud sediment thickness, and sunspots. He found that in each case, the
statistic behaved as 𝑅/𝑆(𝑛) ∝ 𝑛 𝑘 for some k. As discussed below, he estimated k using
linear regression, and we will follow him in denoting this estimate as K. He found that K was
approximately normally distributed with mean 0.72 ± 0.006 . He did acknowledge that “K does
vary slightly with different phenomena”, and that it was “the mean value of a quantity that has
ranged from 0.46 to 0.96” (large for a Gaussian fit), however to a first approximation it
(/)
appeared that the value of 0.72
might hold some global significance.

overlooked. As we shall see, the buscarmen


At this point, it is worth highlighting an aspect of Hurst’s work which often gets
𝑅/𝑆 ú
statistic has enjoyed great use over the past fifty years
(although its use tends now to be deprecated in favour of more accurate ways of estimation).
However, the modern method of estimating the R/S exponent k is not that originally used by
Hurst. His estimate K was obtained by assuming a known constant of proportionality:
specifically he assumed the asymptotic (i.e., for large n) law that𝑅/𝑆(𝑛) = (𝑛/2) 𝑘 . A doubly
logarithmic plot of values of 𝑅/𝑆(𝑛)
against 𝑛/2 should produce a straight line, the slope of
which is taken as K. By assuming a known constant of proportionality, Hurst was effectively
performing a one parameter log-regression to obtain his estimate of k.


His reason for choosing this approach was that it implies 𝑅 / 𝑆 ( 2 ) =exactamente
1 (en
realidad es igual1 /√2pero Hurst estaba calculando las desviaciones estándar de la
población en lugar de las de la muestra, es decir, dividiendo por n en lugar de𝑛 − 1), por lo
que este 'valor computable' podría utilizarse en el procedimiento de estimación. Esta
metodología hoy en día se consideraría correctamente como muy dudosa porque implica
ajustar una relación asintótica ( n grande ) mientras se utiliza un valor fijo pequeño supuesto
para la𝑛=1 punto. Este error lógico fue inmediatamente comentado en el mismo número de
la revista Te Chow [ 23 ]. Como veremos, Mandelbrot introdujo más tarde el método de
estimación ahora estándar eliminando este punto fijo y realizando una regresión logarítmica
de dos parámetros para obtener la pendiente. El método original de Hurst fue olvidado y la
mayoría de los autores desconocen que no era igual al método moderno; de hecho, muchos
citan el resultado de Hurst de 0,72
sin saber que se obtuvo mediante un análisis inadecuado.
A pesar de estas deficiencias, el resultado clave de Hurst de que las estimaciones de k
eran aproximadamente 0,72 probablemente no habría sido digno de mención ni controvertido
en sí mismo si no hubiera demostrado que, utilizando modelos estocásticos
contemporáneos, este comportamiento no podía explicarse.
A principios de la década de 1950, el modelado estocástico de los flujos de los ríos era
inmaduro y, por lo tanto, el único modelo que Hurst podía considerar era el modelo iid
gaussiano de Hazen [ 18 ] y Sudler [ 19 ]. Deducir rigurosamente la distribución del rango bajo
este modelo estaba más allá de las habilidades matemáticas de Hurst, pero al considerar las
asintóticas de un juego de lanzamiento de moneda y apelar al teorema del límite central,
produjo una solución heurística extraordinariamente buena. Su trabajo demostró que, bajo el
supuesto gaussiano independiente, el exponente k debería ser igual a 0,5. En otras palabras,
Hurst había demostrado que los modelos hidrológicos contemporáneos fundamentalmente
no concordaban con la evidencia empírica. Esta discrepancia entre la teoría y la práctica se
conoció como el "fenómeno Hurst". Vale la pena aclarar aquí una posible ambigüedad: desde
que se acuñó la frase, el 'fenómeno de Hurst' se ha atribuido a varios aspectos de series
temporales y/o procesos estocásticos. Para mayor claridad, utilizaremos el término en el
sentido de “la estadística𝑅 / 𝑆 ( crece
𝑛 ) empíricamente más rápido que𝑛1/2 ”._ _
La observación de Hurst desencadenó una serie de investigaciones que finalmente
(/)
condujeron al desarrollo formal de la memoria larga. El propio Hurst no ofreció ninguna

fundamental podría residir en el supuesto de independencia: buscarmen


explicación específica para el efecto, aunque claramente sospechaba que la causa
ú

Aunque en eventos aleatorios sí ocurren grupos de valores altos o bajos, su tendencia a


ocurrir en eventos naturales es mayor. ... No hay una periodicidad evidente, pero hay
largos tramos en los que las inundaciones son generalmente elevadas, y otros en los
que, en general, son escasas. Estos tramos se producen sin ninguna regularidad ni en su
momento de aparición ni en su duración.

([ 14 ], §6)

Después de estas observaciones empíricas cruciales y varias publicaciones de


seguimiento [ 24 , 25 , 26 ], el propio Hurst no jugó ningún papel directo en el desarrollo
matemático a largo plazo de la memoria larga. El propósito específico de su investigación fue
diseñar un sistema para controlar el Nilo con una serie de pequeñas presas y embalses.
Estos planes se convirtieron más tarde en la presa de Asuán, y Hurst siguió actuando como
consultor científico hasta los ochenta años [ 6 ].
2.2. Reacciones al fenómeno Hurst
El hallazgo de Hurst tomó por sorpresa a la comunidad hidrológica, no sólo por el
enigma intrínseco, sino por su importancia potencial. Como se mencionó anteriormente, el
𝑅/𝑆(𝑛) La estadística es una aproximación a la altura ideal de la presa durante n años. Si
había que creer en el hallazgo de Hurst, y 𝑅/𝑆(𝑛) 𝑛
aumentó más rápido que 1/2 ,_ _habría
implicaciones potencialmente importantes para el diseño de la presa. En otras palabras, las
represas diseñadas para horizontes temporales prolongados podrían ser demasiado bajas,
con inundaciones como una posible consecuencia.
Aunque el debate sobre los hallazgos de Hurst, que posteriormente evolucionó hasta
convertirse en el debate sobre la memoria larga, inicialmente se limitó en gran medida a la
comunidad hidrológica, fortuitamente también pasó a la literatura matemática más
convencional, un hecho que sin duda ayudó a elevar su perfil interdisciplinario en años
posteriores. años. A pesar del atractivo matemático poco claro de lo que era esencialmente
un tema de nicho, el eminente probabilista William Feller [ 22 ] contribuyó en gran medida con
la publicación de un breve artículo. Apelando a la teoría del movimiento browniano, demostró
que Hurst tenía razón; para secuencias de variables aleatorias iid estandarizadas con
varianza finita, la distribución asintótica del rango ajustado, 𝑅(𝑛)
, debe obedecer el 1/2 ley:
__ 𝑛
𝔼 [𝑅 (𝑛)] ∼ ( 𝜋2 )1/2.𝑛_ 1/2_Cabe_ _ destacar que Feller estaba estudiando la distribución del
rango ajustado, 𝑅 ( 𝑛, )no el rango ajustado reescalado𝑅 / 𝑆 ( .𝑛La) importancia de dividir por la
desviación estándar no se apreció hasta Mandelbrot; sin embargo, más tarde se demostraría
que los resultados de Feller también son válidos para esta estadística.
Al probar y ampliar (ya que el supuesto de gaussianidad podría debilitarse) el resultado
(/)
de Hurst, Feller logró confirmar que había un fenómeno de interés y también que debería

buscarmen
interesar tanto a los matemáticos como a los hidrólogos. En el transcurso de la década de
ú
1950 se obtuvieron resultados más precisos, aunque lamentablemente se desvió la atención
hacia la consideración del rango simple (por ejemplo, [ 27 ]) en contraposición a 𝑅/𝑆 . La
distribución exacta de𝑅(𝑛) se consideró, en general, intratable; una excepción notable es la
del caso gaussiano iid más simple, donde [ 28 ]

𝜋 1/2 _⎛⎜_ 1 𝑛 − 1 1
𝔼 [ 𝑅 ( 𝑛 ) ] = ( 2 ) ⎜ 𝜋 ∑ −−−−
⎝ 𝑘= 1 √𝑘( 𝑛
Habiendo demostrado de manera concluyente que los hallazgos de Hurst eran realmente
dignos de investigación, se propusieron varias explicaciones posibles para el fenómeno del
mismo nombre. Se supuso que el efecto fue causado por (al menos) una de las siguientes
propiedades del proceso: (a) una distribución marginal "inusual", (b) no estacionariedad, (c)
transitoriedad (es decir, comportamiento preasintótico ), o (d) efectos de autocorrelación a
corto plazo.
Para los datos originales de Hurst, la primera de las soluciones propuestas no era
relevante porque muchos de sus datos eran claramente gaussianos. Moran [ 29 ] afirmó que
el efecto podría explicarse mediante el uso de una secuencia de variables aleatorias iid con
una condición de momento particular en la distribución. Aunque Feller [ 22 ] había
𝑛
demostrado que este caso todavía producía asintóticamente la 1/2 Según _ _ la ley, Moran
demostró que en tales casos la fase transitoria (o preasintótica ) que exhibe el fenómeno de
Hurst podría extenderse arbitrariamente. Moran utilizó una distribución Gamma, aunque para
lograr el efecto la distribución tenía que estar muy sesgada, descartándola así como una
explicación práctica del efecto de Hurst. Además, Moran señaló que si se abandonara por
𝛼
completo el supuesto de varianza finita y en su lugar se aplicara un modelo simétrico Si se
suponía una distribución estable, el fenómeno de Hurst aparentemente podría explicarse:
𝔼 [ 𝑅 ( 𝑛 ) ] ∼ ℓ𝑛 1 < 𝛼 ≤ 2
1/𝛼
, para . y algunos ℓ conocidos (computables) . Sin embargo,
como demostró más tarde Mandelbrot, la división por la desviación estándar es realmente
crucial. En otras palabras, si bien los argumentos de Moran eran correctos, eran irrelevantes
porque el objeto de interés real era el rango ajustado reescalado . Varios estudios de
Montecarlo, en particular los de Mandelbrot y Wallis [ 30 ], confirmaron que para variables
aleatorias iid,𝑅/𝑆(𝑛) 𝑛
sigue asintóticamente una 1/2 ley.
_ _ Las pruebas posteriores de
Mandelbrot [ 31 ] y Mandelbrot y Taqqu [ 32 ] han asegurado el interés matemático en la 𝑅/𝑆
estadística hasta la actualidad. Sin embargo, hay una sutileza: en el caso de variables
𝑛 𝑅/𝑆 (𝑛)
aleatorias iid con varianza finita, − 1/2 _ _ converge en distribución a una función del
𝑛 𝑅/𝑆 (𝑛)
puente browniano, mientras que en un caso estable − 1/2 _ _ converge en distribución a
una función de una medida aleatoria de Poisson, como lo analiza Samorodnitsky [ 10 ]. La
(/)
normalización es la misma, pero el comportamiento limitante es diferente.

buscarmen
La segunda posible explicación del fenómeno Hurst, la no estacionariedad, es más difícil
de descartar y es más una cuestión filosófica (y física) que matemática. ú

¿Tiene sentido hablar de una media invariante en el tiempo durante miles de años? Si se
dispusiera de realizaciones lo suficientemente largas de tales series temporales, ¿serían
realmente estacionarias?

([ 33 ], §3.2)

Una vez que asumimos la posibilidad de la no estacionariedad, no es difícil imaginar que


esto podría conducir a una explicación del fenómeno. De hecho, el propio Hurst [ 34 ] sugirió
que la no estacionariedad podría ser una explicación; sin embargo, su heurística que
involucra una baraja de naipes estaba lejos de ser matemáticamente formalizable. Klemeš [
35 ] y Potter [ 36 ] proporcionaron más tarde más evidencia; sin embargo, el primer modelo
matemático viable riguroso fue el de Bhattacharya et al. [ 37 ], en el que los autores
demostraron que un proceso de memoria corta perturbado por una tendencia monótona no
lineal puede exhibir el fenómeno de Hurst. Su estudio muestra por qué es crucial distinguir
entre el "fenómeno Hurst" y la "memoria larga". El proceso descrito por Bhattacharya et al. no
tiene memoria larga pero exhibe el fenómeno de Hurst (recuerde nuestra definición
específica de este término).
En su influyente artículo, Klemeš [ 35 ] no sólo demostró que el fenómeno de Hurst
podría explicarse por la no estacionariedad, sino que argumentó que asumir la
estacionariedad puede estar infundado:

La cuestión de si los procesos naturales son estacionarios o no es probablemente


filosófica. ... probablemente no existe una sola serie de tiempo histórica de la cual las
matemáticas puedan decir con certeza si es estacionaria o no... Tradicionalmente, se ha
asumido que, en general, los procesos geofísicos, biológicos, económicos y otros
procesos naturales no son estacionarios, pero dentro de Los lapsos de tiempo
relativamente cortos pueden aproximarse bien mediante modelos estacionarios.

[ 35 ]

Como ejemplo, Klemeš sugirió que un gran terremoto podría afectar drásticamente a una
cuenca fluvial hasta el punto de inducir un cambio de régimen (es decir, un elemento de no
estacionariedad). Sin embargo, a una escala mayor (espacial y temporal), el terremoto y su
deformación local de la Tierra pueden verse como parte de un “modelo de la Tierra”
estacionario general. Por tanto, elegir entre las dos formas es, hasta cierto punto, una
cuestión de creencia personal. De hecho, Mandelbrot consideró (y publicó) otros modelos
con un tipo particular de conmutación no estacionaria, incluso mientras formulaba su
(/)
modelo estacionario FGN, pero desafortunadamente Klemeš no estaba al tanto de ese

buscarmen
trabajo, sobre el cual tal vez podría haber ocurrido una discusión más fructífera.
Si descartamos esta explicación y asumimos la estacionariedad, debemos recurrir ú
a la
tercera y cuarta explicaciones posibles, a saber, la fugacidad (es decir, el comportamiento
preasintótico) y/o la falta de independencia. Estos dos efectos están relacionados: es
probable que los efectos de autocorrelación a corto plazo introduzcan comportamientos
preasintóticos significativos. Como se mencionó anteriormente, el propio Hurst sugirió que
algún tipo de dependencia en serie podría explicar el efecto, y Feller sugirió:

Es concebible que el fenómeno [de Hurst] pueda explicarse probabilísticamente,


partiendo del supuesto de que las {𝑋𝑘 }no son independientes...
variables
Matemáticamente, esto requeriría tratar las variables{𝑋𝑘 }como un proceso de Markov.

[ 22 ]

Pronto, sin embargo, Barnard [ 38 ] afirmó haber demostrado que los modelos
𝑛
markovianos todavía conducían a la 1/2 ley
_ _ y se demostraría más adelante [ 31 , 39 ] que
cualquier forma de autocorrelación entonces conocida debe conducir asintóticamente al
mismo resultado. La condición requerida en la función de autocorrelación resultó ser que sea
𝜌 ( ·[)40 ]:
sumable, por lo que para variables aleatorias iid con ACF

𝜋 1/2 ⎛_ _ ∞
𝔼 [𝑅/𝑆 (𝑛)] ∼(2) ⎜ ∑ 𝜌 ( (1)
⎝𝑘= − ∞
Incluso antes de que esto se demostrara formalmente, se sabía en general que sería
necesaria alguna estructura de autocorrelación complicada para explicar el fenómeno de
Hurst:

Se ha sugerido que la correlación o dependencia serial [podría causar el fenómeno


Hurst]. Sin embargo, esto no puede ser cierto a menos que la dependencia serial sea de
un tipo muy peculiar, pues en todos los modelos plausibles de dependencia serial la
serie de valores siempre se aproxima mediante un [movimiento browniano] cuando la
escala de tiempo es suficientemente grande. Una teoría más plausible es que las series
experimentales utilizadas por Hurst, como resultado de la correlación serial, no son lo
suficientemente largas para que la fórmula asintótica sea válida.

[ 20 ]

Así, Moran sostenía que, dado que ninguna estructura de autocorrelación “razonable”
podía explicar el fenómeno de Hurst, se debería suponer que el efecto observado fue
causado por un comportamiento preasintótico, cuyo alcance estaba influenciado por alguna
(/)
forma de dependencia local. . En otras palabras, estaba sugiriendo que un proceso de

buscarmen
memoria corta podría explicar el fenómeno Hurst en escalas de tiempo observadas.
ú
Esta cuestión tiene una importancia tanto práctica como filosófica. Más tarde algunos
argumentarían que, independientemente del modelo “verdadero”, cualquier proceso que
pudiera exhibir el fenómeno de Hurst en las escalas de tiempo observadas (o requeridas)
sería suficiente para fines prácticos. El uso de tales procesos requiere una elección. Se
podría aceptar el fenómeno Hurst como genuino y reconocer que, aunque teóricamente
incorrecto, tal modelo es suficientemente bueno para el propósito deseado. Alternativamente,
se podría rechazar el fenómeno de Hurst simplemente como un efecto transitorio
preasintótico y, por lo tanto, cualquier modelo que replique el efecto en rangos observados de
n es potencialmente válido. Mandelbrot, por su parte, fue muy crítico con quienes seguían
este último enfoque:

Hasta ahora, tal convergencia [hacia el 𝑛1/2 ley]


_ _ nunca se ha observado en hidrología. Así,
aquellos que consideran que el efecto de Hurst es transitorio implícitamente atribuyen
una importancia inmerecida al valor [del tamaño de la muestra]... Estos académicos se
condenan a sí mismos a nunca ser testigos del desarrollo asintótico completo de los
modelos que postulan.

[ 39 ]

A pesar de esto, el concepto de fugacidad inducida por la memoria corta fue explorado
tanto antes como después del trabajo de Mandelbrot. Matalas y Huzzen [ 41 ] realizaron un
(1)
riguroso análisis de Monte Carlo del AR modelo y demostró que para una correlación serial
media n y alta de retraso uno, se podría inducir el fenómeno de Hurst (aunque Matalas y
Huzzen en realidad estaban usando la estimación K errónea original de Hurst ). Fiering [ 42 ]
logró construir un modelo más sofisticado; sin embargo, descubrió que necesitaba usar un
AR( 20 )
proceso para inducir el efecto: un número irrealmente grande de retrasos que sean
útiles para el modelado.
En resumen, a principios de la década de 1960, más de una década después de los
descubrimientos originales de Hurst, todavía no se había encontrado ninguna explicación
satisfactoria para el fenómeno Hurst. Para citar [ 35 ] nuevamente:

Desde que Hurst publicó sus famosos argumentos para algunas series temporales
geofísicas... el ya clásico fenómeno de Hurst ha seguido atormentando a estadísticos e
hidrólogos. Para algunos, se ha convertido en un rompecabezas que hay que explicar,
para otros en una característica que sus modelos deben reproducir y, para otros, en un
fantasma que hay que conjurar.

Fue en ese momento cuando Benoît Mandelbrot se enteró del fenómeno.


3. Modelos fraccionarios de Mandelbrot
(/)
A principios de la década de 1960, Mandelbrot había trabajado intensamente en el

buscarmen
floreciente tema de las finanzas matemáticas y el problema de modelar cantidades como los
ú
precios de las acciones. Un elemento central de este tema fue la "hipótesis del paseo
aleatorio" que proporcionó modelos de movimiento browniano. Esto fue propuesto
implícitamente por primera vez en la tesis doctoral fundamental (aún no descubierta hace
mucho tiempo) de Bachelier [ 43 ]. El desarrollo detallado de este tema también es
interesante pero está fuera del alcance de este artículo. Baste decir aquí que, aunque el
modelo de Bachelier fue reconocido como un modelo de trabajo adecuado que parecía
ajustarse tanto a la intuición como a los datos, también podría beneficiarse de mejoras. Se
propusieron varias modificaciones, pero una característica común que todas compartían era
el supuesto gaussiano subyacente.
En un artículo innovador, Mandelbrot [ 44 ] propuso abandonar el supuesto de
gaussianidad y asumir en su lugar una distribución de cola pesada, específicamente la
𝛼
distribución simétrica. -distribución estable (por ejemplo, [ 45 ], §1.1). En resumen, esta
noción fue muy controvertida; por ejemplo, véase Cootner [ 46 ]. Sin embargo, el artículo fue
significativo por dos razones. En primer lugar, ayudó a dar credibilidad al creciente estudio de
las distribuciones de colas pesadas y los procesos estocásticos. En segundo lugar, era
𝛼
indicativo de la fascinación de Mandelbrot por la escala matemática. El Las distribuciones
estables tienen la atractiva propiedad de que una suma adecuadamente reponderada de
𝛼
tales variables aleatorias es en sí misma una -variable aleatoria estable. Esta pasión por la
escala permanecería con Mandelbrot durante toda su vida y se resume en su famosa
geometría fractal.
Volviendo a los resultados de Hurst, la familiaridad de Mandelbrot con la escala le ayudó
a reconocer inmediatamente el fenómeno de Hurst como sintomático de esto y, como relató
más tarde [ 6 , 47 ], asumió que podría explicarse por procesos de cola pesada. Por lo tanto,
se sorprendió cuando se dio cuenta de que los datos de Hurst no sólo eran esencialmente
gaussianos, sino que, como se analizó anteriormente, el rango ajustado reescalado no es
sensible a la distribución marginal. En cambio, se dio cuenta de que sería necesario un nuevo
enfoque. De acuerdo con la idea de escala, introdujo el término “autosimilar” e introdujo
formalmente el concepto en su forma moderna: Dejemos que 𝑌( 𝑡 )
ser un proceso estocástico
de tiempo continuo. Entonces 𝑌( 𝑡 )
se dice que es autosimilar, con parámetro de
autosemejanza H , si para todo c positivo , 𝑌( 𝑐 𝑡 ) =𝑑 𝑐 𝑌( 𝑡 )
𝐻 . Utilizando este concepto,
Mandelbrot [ 48 ], sentó las bases de los procesos que inicialmente se convertirían en los
modelos paradigmáticos en el campo de la memoria larga, el modelo autosimilar de
movimiento browniano fraccional (FBM) y sus incrementos, el modelo gaussiano fraccional
dependiente de largo alcance modelo de ruido (FGN). Más tarde, Mandelbrot lamentó el
término “autosimilar” y llegó a preferir “autoafín”, porque las escalas en el tiempo y el espacio
no eran necesariamente las mismas, pero la terminología revisada nunca tuvo la misma
(/)
aceptación.

buscarmen
Llegados a este punto, es necesario describir informalmente la FBM. Es un proceso
ú
gaussiano de tiempo continuo y es una generalización del movimiento browniano ordinario,
con un parámetro adicional h . Este parámetro puede oscilar entre cero y uno (no inclusivo
para evitar patologías) con diferentes valores que proporcionan tipos de comportamiento
cualitativamente diferentes. El caso ℎ = 1/2 _ _
Corresponde al movimiento browniano estándar.
We remark that the naming and notation of the parameter h has been a source of
immense confusion over the past half-century, with various misleading expressions such as
the “Hurst parameter”, the “Hurst exponent”, the “Hurst coefficient”, the “self-similarity
parameter” and the “long memory parameter”. Moreover, the more traditional notation of an
upper-case H does not help since it disobeys the convention of using separate cases for
constants (parameters) and random variables (statistics). For clarity, in what follows, we will
reserve the notation h simply to denote the “fractional Brownian motion parameter”, and will
distinguish it from experimentally obtained estimators such as Hurst’s K or Mandelbrot’s
estimator J, and the self-similarity parameter H. In certain cases, all these will be the same,
but we wish to allow for the possibility that they will not be.
Fractional Brownian motion can be thought of in several different and equivalent ways,
for example as a fractional derivative of standard Brownian motion, or as stochastic integral.
These details need not concern us here; the most important fact is that FBM is exactly self-
similar, which means that a “slowed-down” version of a process will, after a suitable spatial re-
scaling, look statistically identical to the original, i.e., they will have the same finite
dimensional distributions. In this sense, FBM, like standard Brownian motion (which of course
is just a special case of FBM), has no characteristic time-scale, or “tick”.
In practical applications, it is necessary to use a modification of FBM because it is (like
standard Brownian motion) a continuous time process and non-stationary. Thus, the
increments of FBM are considered; these form a discrete process which can be studied using
conventional time series analysis tools. These increments, called fractional Gaussian noise
(FGN), can be considered to be the discrete approximation to the “derivative” of FBM. Note
that in the case of ℎ = 1/2
, FGN is simply the increments of standard Brownian motion, i.e.,
white noise. (Mandelbrot and Van Ness [49], corollary 3.6) showed that this process is
stationary, but most importantly (for its relevance here), it exhibits the Hurst phenomenon: for
some 𝑐 > 0 𝑅/𝑆(𝑛) ∼ 𝑐𝑛
, ℎ . This result was immensely significant; it was the first time since
Hurst had first identified the phenomenon, that anyone had been able to exhibit a stationary,
Gaussian stochastic process capable of reproducing the effect. The mystery had been
partially solved; there was such a process, and for over a decade it remained the only model
known to be able to fully explain the Hurst phenomenon.
Mandelbrot [48] then proceeded to show that such a process must have a spectral
density function that blows up at the origin. By proposing such a model, he realised he was
attempting to explain with one parameter h both low- and high-frequency effects, i.e., he was
“… postulating the same mechanism for the slow variations of climate and for the rapid
(/)
variations of precipitation”. He also recognised that the auto-correlation function of the
1/2 < ℎ < 1
increments would decay slower than exponentially, and (for
buscarmen ) would not be
ú
summable. This correlation structure, which is now often taken to be the definition of long
memory itself, horrified some. Concurrently, the simplicity of FGN, possessing only one
parameter h, concerned others. We shall consider these issues in depth later, but the key
point was that although Mandelbrot had ‘conquered’ the problem, to many it was somewhat
of a Pyrrhic victory [35].
3.1. Initial Studies of Mandelbrot’s Model
Mandelbrot immediately attempted to expand on the subject although his papers took
time to get accepted. He ultimately published a series of five papers in 1968–1969 through
collaborations with the mathematician John Van Ness and the hydrologist James Wallis.
Taken as a whole, these papers offered a comprehensive study of long memory and fractional
Brownian motion. They helped publicise the subject within the scientific community and
started the debates about the existence of long memory and the practicality of FGN, which
have continued until the present day.
The first of these papers [49] formally introduced FBM and FGN and derived many of
their properties and representations. The aim of this paper was simply to introduce these
processes and to demonstrate that they could provide an explanation for the Hurst
phenomenon; this was succinctly stated:

We believe that FBMs do provide useful models for a host of natural time series and
wish therefore to present their curious properties to scientists, engineers and
statisticians.

Mandelbrot and Van Ness argued that all processes thus far considered have “the
property that sufficiently distant samples of these functions are independent, or nearly so”,
yet in contrast, they pointed out that FGN has the property “that the ‘span of interdependence’
between [its] increments can be said to be infinite”. This was a qualitative statement of the
difference between short and long memory and soon led to the formal definition of long
memory. As motivation for their work, they cited various examples of observed time series
which appeared to possess this property: in economics [50,51], “ 1/𝑓 noises” in the
fluctuations of solids [52], and hydrology [14].
Intriguingly, and undoubtedly linked to Mandelbrot’s original interest in heavy-tailed
processes, (Mandelbrot and Van Ness [49], §3.2) noted:
(/)
If the requirement of continuity is abandoned, many other interesting self-similar

buscarmen
processes suggest themselves. One may for example replace [the Brownian motion] by
𝛼 ú
a non-Gaussian process whose increments are [ -] stable … Such increments necessarily
have an infinite variance. “Fractional Lévy-stable random functions” have moreover an
infinite span of interdependence.
In other words, the authors postulated a heavy-tailed, long memory process. It would be
over a decade before such processes were properly considered due to difficulties arising from
the lack of formal correlation structure in the presence of infinite variance. However, a
preliminary demonstration of the robustness of R/S as a measure of LRD was given in [30],
using a heavy tailed modification of fBm which the authors dubbed “fractional hyperbolic
motion”.
One key point which is often overlooked is that Mandelbrot and Van Ness did not claim
that FGN is necessary to explain the Hurst phenomenon: “… we selected FBM so as to be able
to derive the results of practical interest with a minimum of mathematical difficulty”. Often,
Mandelbrot was incorrectly portrayed as insisting that his, and only his, model solved the
problem. Indeed, Mandelbrot himself took an interest in alternative models [52], although as
we will later see, he essentially rejected Granger and Hosking’s ARFIMA which was to become
the standard replacement of FGN in statistics and econometrics literatures.
Furthermore, neither did the authors claim that they were the first to discover FBM. They
acknowledged that others (e.g., [53]) had implicitly studied it; however, Mandelbrot and Van
Ness were undoubtedly the first to attempt to use it in a practical way.
Having ‘solved’ Hurst’s riddle with his stationary fractional Gaussian model, Mandelbrot
was determined to get FGN and FBM studied and accepted, in particular by the community
which had most interest in the phenomenon, hydrology. Therefore, his remaining four
important papers were published in the leading hydrological journal Water Resources
Research. These papers represented a comprehensive study of FBM in an applied setting, and
were bold; they called for little short of a revolution in stochastic modelling:

... current models of statistical hydrology cannot account for either [Noah or Joseph]
effect and must therefore be superseded. As a replacement, the ‘self-similar’ models
that we propose appear very promising.

[39]

As its title suggests, Mandelbrot and Wallis [39] introduced the colourful terms “Noah
Effect” and “Joseph Effect” for heavy tails and long memory respectively; both labels
referencing key events of the Biblical characters’ lives. Ironically, the river level data were, in
fact, close enough to Gaussian to dispense with the “Noah Effect” so the actual content of the
paper was largely concerned with the “Joseph Effect”, but rainfall itself provides a rich source
of heavy tailed, “Noah” datasets. However, Mandelbrot preferred treating these two effects
(/)
together as different forms of scaling; spatial in the former and temporal in the latter.

buscarmen
Mandelbrot and Wallis [39] defined the “Brownian domain of attraction” (BDoA) and
ú
showed that such BDoA cannot account for either effect and should therefore be discarded.
The BDoA was (rather loosely) defined as the set of discrete-time stochastic processes which
obey three conditions; namely, the Law of Large Numbers, the Central Limit Theorem, and
asymptotic independence of past and future partial sums. Alternatively, the BDoA is the set of
processes which are either asymptotically Brownian, or can be well-approximated by
Brownian motion. A process in the BDoA is, in some sense, “nice”, i.e., it is Gaussian or
Gaussian-like and has short memory, and was given the term “smooth”. Processes outside of
the BDoA were labelled “erratic” . This “erratic” behaviour could be caused by one, or both, of
the Joseph and Noah effects. Mandelbrot and Wallis showed that processes lying within the
𝑛
BDoA will, after an initial transient behaviour, obey the 1/2 law. They rejected, on
philosophical grounds, the idea that the Hurst phenomenon might be caused by transient
effects. Mandelbrot later preferred the terms “mild” to “nice”, and subdivided “erratic” into
heavy tailed “wild” and strongly dependent “slow”. We stick with his original terminology.
Mandelbrot procedió a aportar más pruebas en apoyo de su modelo. Mandelbrot y Wallis
[ 15 ] incluyeron varios gráficos de muestra de realizaciones simuladas de FGN con h
variables . El objetivo explícito era "fomentar la comparación de [la] serie artificial con el
registro natural que interesa al lector". Estas simulaciones se realizaron utilizando uno de los
dos métodos desarrollados por los autores, que eran diferentes tipos de aproximaciones
truncadas. Como lo documentó Mandelbrot [ 54 ], pronto se descubrió que una de sus
aproximaciones estaba lejos de ser adecuada porque no lograba reproducir con precisión los
efectos deseados. Los algoritmos también fueron lentos de implementar; un problema
práctico importante cuando el tiempo de la computadora era costoso y la potencia de
procesamiento limitada. Mandelbrot [ 54 ] introdujo por tanto un algoritmo más eficiente.
Más tarde, se crearía un algoritmo exacto [ 55 , 56 ] que forma la base de los algoritmos
modernos [ 1 , 57 ] que utilizan la Transformada Rápida de Fourier. Mandelbrot y Wallis
quisieron someter sus simulaciones a 𝑅/𝑆
análisis, pero reconocieron el defecto lógico
mencionado anteriormente en el enfoque de Hurst. Por lo tanto, desarrollaron una regresión
logarítmica sistemática de dos parámetros para obtener una estimación del parámetro h de
FGN utilizando el exponente R/S. Desde entonces, este enfoque se ha convertido en el
método estándar para estimar el exponente R/S. Siguiendo la recomendación de Mandelbrot
en sus volúmenes “Selecta”, usaremos J para denotar el exponente estimado por este
método.
Los caminos de muestra simulados fueron sometidos a ambos 𝑅/𝑆
y análisis espectral, y
para ambos casos se encontró que los caminos simulados coincidían en gran medida con la
teoría, es decir, los caminos de muestra parecían representaciones suficientemente buenas
de los procesos teóricos. Para el 𝑅/𝑆
análisis, se encontró, como se esperaba, que existían
tres regiones distintas: comportamiento transitorio, comportamiento 'Hurst' y
comportamiento asintótico "1/2". Esta última región se debió en su totalidad a que las
(/)
simulaciones eran esencialmente aproximaciones de memoria corta a los procesos de

buscarmen
memoria larga; las medias móviles infinitas se truncaron a finitas. Por lo tanto, esta tercera
región podría eliminarse mediante una síntesis cuidadosa, es decir, haciendo que úlos
promedios móviles sean mucho más largos que la longitud del conjunto. Además, más tarde
se demostró que la región transitoria [ 58 ] era en gran medida una característica de un error
de programación.
Mandelbrot y Wallis [ 59 ] aplicaron sus𝑅/𝑆
método para muchos de los mismos tipos de
datos que Hurst [ 14 , 24 ] y Hurst et al. [ 26 ], y de manera similar encontró evidencia
significativa a favor de la hipótesis de la memoria larga. En una comparación de K de Hurst
con su J , Mandelbrot y Wallis señalaron que K tenderá a subestimar h cuando ℎ > 0,72 pero
sobreestima cuando ℎ < 0,72 . Así, el célebre hallazgo de Hurst de un promedio global de 0,72
estuvo fuertemente influido por su deficiente método, y su desviación estándar estimada
respecto de esta media fue subestimada. Hace tiempo que se ha olvidado este importante
punto: que los hallazgos empíricos originales que ayudaron a generar el tema de la memoria
larga eran sistemáticamente defectuosos.
A continuación, Mandelbrot y Wallis [ 30 ] emprendieron un estudio detallado de Monte
Carlo sobre la robustez a la no gaussianidad de sus 𝑅/𝑆
método. Como se mencionó
anteriormente, en general 𝑅/𝑆
demostró ser muy robusto. Las diferentes distribuciones
estudiadas fueron gaussianas, lognormales, “hiperbólicas” (una distribución sesgada de cola
𝛼
pesada, no -estable pero atraído por esa ley), y gaussiano truncado (para lograr una curtosis
menor que la gaussiana). La distribución del rango ajustado no normalizado, 𝑅(𝑛) , demostró
ser altamente dependiente de la distribución; sin embargo, la división por𝑆 ( 𝑛corregido
) por
esto. Para cualquier secuencia de variables aleatorias iid, su J estimado siempre fue
(cercano a)1/2._ _
Al estudiar los casos dependientes, consideraron varias transformaciones no lineales
(como transformaciones polinómicas o exponenciales) y descubrieron que la robustez aún
se mantenía. Sin embargo, 𝑅/𝑆 se mostró susceptible en presencia de fuertes periodicidades;
un hecho descartado con bastante optimismo: “Los componentes cíclicos agudos rara vez
ocurren en los registros naturales. Es más probable encontrar mezclas de ondas que tengan
longitudes ligeramente diferentes…”.
Finalmente, Mandelbrot y Wallis [ 30 ] intrigantemente reemplazaron las variables
gaussianas en su simulador FGN con variables "hiperbólicas". Aunque ahora se sabe que
tiene inconvenientes, durante mucho tiempo este fue el único intento de simular un proceso
de memoria larga y de cola pesada.
3.2. Reacciones al modelo de Mandelbrot
By proposing heavy tailed models to economists, Mandelbrot had had a tough time
advocating against orthodoxy [7]. Because his fractional models were similarly unorthodox,
he learned from his previous experience, and was more careful about introducing them to
hydrologists. By producing several detailed papers covering different aspects of FBM, he had
(/)
covered himself against charges of inadequate exposition. Unsurprisingly however, many

buscarmen
hydrologists were unwilling to accept the full implications of his papers.
ú
Firstly, Mandelbrot’s insistence on self-similar models seemed somewhat implausible
and restrictive, and seemed to totally ignore short-term effects. Secondly, Mandelbrot’s model
was continuous-time which, although necessary to cope with self-similarity, was only useful in
a theoretical context because we live in a digital world; data are discrete and so are
computers. Mandelbrot was primarily interested in FBM; he saw the necessary discretisation,
FGN, as its derivative, both literally and metaphorically. As soon as his models were applied to
the real world, they became compromised:

The theory of fractional noise is complicated by the motivating assumptions being in


continuous time and the realizable version being needed in discrete time.

([60], §6.2)

In one major respect, Mandelbrot was simply unlucky with timing. Soon after his papers
about FBM were published, the hugely influential book by Box and Jenkins [61] was published,
revolutionising the modelling of discrete time series in many subject areas.
Prior to 1970, multiple-lag auto-regressive or moving average models had been used (and
as previously mentioned had failed to adequately replicate the Hurst phenomenon), but the
Box–Jenkins models combined these concepts, together with an integer differencing
parameter d, to produce the very flexible class of ARIMA (𝑝,𝑑,𝑞)
models. As in other scientific
fields, many hydrologists were attracted to these models, and sought to explore the
possibility of using them to replicate the Hurst phenomenon.
It is important to note that ARIMA models cannot genuinely reproduce the asymptotic
Hurst phenomenon since all ARIMA models either have short memory, or are non-stationary.
However, by choosing parameters carefully, it can be shown that it is possible to replicate the
observed Hurst phenomenon over a large range of n. O’Connell [33] was an early exponent of
this idea; specifically, he used an ARMA (1,1) model which could (roughly) preserve a given
first-lag auto-correlation as well as h. For completeness, we mention that other modelling
approaches were investigated to try and replicate the Hurst phenomenon. One such model
was the so-called “broken-line” process detailed by Rodriguez-Iturbe et al. [62], Garcia et al.
[63], and Mejia et al. [64,65] which sought to preserve a twice differentiable spectrum. This
was criticised by Mandelbrot [66] and did not prosper.
To summarise, in the early 1970s, there were two distinct approaches to modelling
hydrological processes. One could use traditional AR processes (or their more advanced
ARMA cousins) which, although able to partially replicate the Hurst phenomenon, were
essentially short memory models. Alternatively, one could use Mandelbrot’s FGN process in
order to replicate the Hurst phenomenon accurately. Unfortunately, this dichotomy was strong
and the choice of approach largely came down to whether accounting for low- or high-
frequency effects was the principal aim for the modeller. Mandelbrot himself was well aware
(/)
(c.f. [39], p. 911) that he was suggesting switching the usual order of priority when modelling

buscarmen
stochastic processes. Many were uncomfortable with this approach because, whereas the
ú
ARMA models could be coerced into replicating the Hurst phenomenon, FGN was completely
uncustomisable with regards to high frequencies.

It remains for the hydrologist to decide which type of behaviour [low- or high-frequency]
is the more important to reproduce for any particular problem. No doubt derivations of
FGN’s preserving both high and low frequency effects will eventually emerge and such a
choice will not be necessary.

([33], §2.3)
Further studies involving ARMA processes were undertaken by Wallis and O’Connell [67],
Lettenmaier and Burges [68] (who proposed a mixture of an ARMA(1,1) model with an
independent AR(1) model), and the set of papers by McLeod and Hipel [69] and Hipel and
McLeod [55,56]. These latter authors were the first to apply to long memory processes the full
Box–Jenkins philosophy of time series estimation: model identification, parameter
estimation, and model-checking. To compare models, they were the first to use formal
procedures such as information criteria, and formally test residuals for whiteness. With this
setup they fitted models to six long-run geophysical time series suspected of possessing
long memory, and found that in each case the best fitting ARMA models were chosen in
preference to FGN. They also fitted more complex ARMA models (than ARMA(1,1)) and
showed again that the observed Hurst statistic can be maintained over the length of series
used. As an aside, the set of papers by McLeod and Hipel were also remarkable for two other
reasons. As mentioned previously, they developed an exact FGN simulator (using the
Cholesky decomposition method), which although computationally expensive, was the first
time anyone had been able to simulate genuine long memory data. Secondly, the authors
derived a maximum likelihood estimator for the FGN parameter h. This was the first proper
attempt at parametric modelling of FGN. Mandelbrot and Taqqu [32] were dismissive of this
approach due to the strong assumptions needed, however from a theoretical statistical point
of view it was a clear breakthrough.
Along with their practical difficulty, another ground for rejecting Mandelbrot’s models was
his sweeping assertions about their physical interpretation. Slightly paraphrasing, he claimed
that, since long memory was the only explanation for the Hurst phenomenon, the underlying
physical processes must possess long memory. This approach of inferring physics from an
empirical model was generally rejected. For a start, many were reluctant to drop the natural
Markovian assumption about nature:
(/)
The past influences the future only through its effect on the present, and thus once a

buscarmenú
certain state of the process has been reached, it matters little for the future
development how it was arrived at.

[35]
Indeed, the renowned hydrologist Vit Klemeš was a leading opponent of Mandelbrot’s
interpretation. As indicated earlier, he personally suspected non-stationarity might be the true
cause for the observed Hurst phenomenon. Whilst he was convinced of the importance of the
Hurst effect, and accepted FGN as an empirical model (he used the phrase “operational
model”) he strongly rejected using it to gain an understanding of physics:

An ability to simulate, and even successfully predict, a specific phenomenon does not
necessarily imply an ability to explain it correctly. A highly successful operational model
may turn out to be totally unacceptable from the physical point of view.

[35]

He likened the apparent success of Mandelbrot’s FGN in explaining the Hurst


phenomenon to the detrimental effect that the Ptolemaic planetary model had on the
development of astronomy. Klemeš had strong reservations about the concept of long
memory, asking:

By what sort of physical mechanism can the influence of, say, the mean temperature of
this year at a particular geographic location be transmitted over decades and centuries?
What kind of a mechanism is it that has carried the impact of the economic crisis of the
1930s through World War II and the boom of the 1950s all the way into our times and
will carry it far beyond?

though he conceded that there were in fact possible mechanisms in the man-made world,
although not, in his view, in the physical one.
More than 20 years later, interviewed by physicist Bernard Sapoval for the online Web of
Stories project, Mandelbrot was to give an answer to Klemeš’ criticism, showing the influence
of subsequent work in physics on critical phenomena on his worldview:
(/)
The consequences of this fundamental idea are hard to accept ... [a]nd many people in

buscarmen
many contexts have been arguing strongly against it, ... If infinite dependence is
ú
necessary it does not mean that IBM’s details of ten years ago influence IBM today,
because there’s no mechanism within IBM for this dependence. However, IBM is not
alone. The River Nile is [not] alone. They’re just one-dimensional corners of immensely
big systems. The behaviour of IBM stock ten years ago does not influence its stock
today through IBM, but IBM the enormous corporation has changed the environment
very strongly. The way its price varied, went up, or went up and fluctuated, had
discontinuities, had effects upon all kinds of other quantities, and they in turn affect us.
And so my argument has always [sic] been that each of these causal chains is totally
incomprehensible in detail, [and] probably exponentially decaying. There are so many of
them that a very strong dependence may be perfectly compatible. Now I would like to
mention that this is precisely the reason why infinite dependence exists, for example, in
physics, in a magnet-because [although] two parts far away have very minor dependence
along any path of actual dependence, there are so many different paths that they all
combine to create a global structure.
Mandelbrot’s esprit d’escalier notwithstanding, Klemeš’ paper remains very worthwhile
reading even today. It also showed how at least two other classes of model could exhibit the
𝜙
Hurst effect, (i) integrated processes, such as random walks, or AR(1) with a parameter
close to 1; and (ii) models of the alternating renewal type with heavy-tailed distributions of the
times between changes in the mean. Ironically, these last renewal models were very similar in
spirit to a model that Mandelbrot had discussed almost 10 years earlier [52,70]. We refer the
reader to a recent historical investigation [71] of these neglected papers, which were many
years ahead of their time.
Klemeš was not alone in his concern over the interpretation of Mandelbrot’s models:

Using self-similarity (with ℎ ≠ 1/2) to extrapolate the correlated behaviour from a finite
time span to an asymptotically infinite one is physically completely unjustified.
Furthermore, using self-similarity to intrapolate [sic] to a very short time span … is
physically absurd.

[72]

Interestingly, in his reply, Mandelbrot [73] somewhat missed the point:


(/)
[The] self-similar model is the only model that predicts for the rescaled range statistic …

buscarmen
precisely the same behaviour as Harold Edwin Hurst has observed empirically. To
ú
achieve the same agreement with other models, large numbers of ad hoc parameters are
required. Thus the model’s justification is empirical, as is ultimately the case for any
model of nature.
Yet another argument used to oppose the use of long memory models arose from a
debate about their practical value. By not incorporating long memory into models, at how
much of a disadvantage was the modeller? Clearly, this is a context-specific question, but the
pertinent question in hydrology is: by how much does incorporating long memory into the
stochastic model change the ideal dam height? One view, shared by Mandelbrot:

The preservation within synthetic sequences … [of h] is of prime importance to engineers


since it characterizes long term storage behaviour. The use of synthetic sequences
which fail to preserve this parameter usually leads to underestimation of long term
storage requirements.

[33]

By ignoring the Hurst phenomenon, we would generally expect to underestimate the ideal
dam height but how quantifiable is the effect? Wallis and Matalas [74] were the first to
demonstrate explicitly that the choice of model did indeed affect the outcome: by comparing
AR(1) and FGN using the Sequential Peak algorithm—a deterministic method of assessing
storage requirements based on the work of Rippl [16] and further developed in the 1960s.
Wallis and Matalas showed that the height depends on both the short and long memory
behaviours, and in general, FGN models require larger storage requirements, as expected.
Lettenmaier and Burges [68] went into more detail by looking at the distribution of the ideal
dam height (rather than simply the mean value) and found it followed extreme value theory
distributions. Lettenmaier and Burges also showed that long memory inputs required slightly
more storage, thus confirming the perception that long memory models need to be used to
guard against ‘failure’.
However Klemeš et al. [75] argued against using this philosophy; instead suggesting that
“failure” is not an absolute term. In the context of hydrology, “failure” would mean being
unable to provide a large enough water supply; yet clearly a minimal deficit over a few days is
a different severity to a substantial drought over many years. Any “reasonable” economic
analysis should take this into account. Klemeš et al. [75] claimed that the incorporation of
long memory into models used to derive the optimum storage height is essentially a “safety
factor”, increasing the height by a few percent, however “… in most practical cases this factor
will be much smaller than the accuracy with which the performance reliability can be
assessed.”
In summary therefore, Mandelbrot’s work was controversial because, although it provided
(/)
an explanation of Hurst’s observations, the physical interpretation of the solution was

buscarmen
unpalatable. There was no consensus regarding the whole philosophy of hydrological
ú
modelling; should the Hurst phenomenon be accounted for, and if so implicitly or explicitly?
Moreover, the new concept of long memory, borne out of the solution to the riddle, was both
non-intuitive and mathematically unappealing at the time.
Much of the debate outlined above was confined to the hydrological community, in
particular the pages of Water Resources Research. With the exception of articles appearing in
probability journals concerning the distributions of various quantities related to the rescaled
adjusted range, little else was known about long memory by statisticians.This was rectified by
a major review paper by Lawrance and Kottegoda [60] which helped bring the attention of the
Hurst phenomenon to the wider statistical community.
One of those non-hydrologists who took up the Hurst “riddle” was the eminent
econometrician Clive Granger. In an almost-throwaway comment at the end of a paper,
Granger [76] floated the idea of “fractionally differencing” a time series, whose spectrum has
a pole at the origin. The ubiquity of 1/𝑓
spectra had been a puzzle to physicists since the work
of Schottky in 1918. Adenstedt [77], derived some properties of such processes but his work
went largely unnoticed until the late 1980s, while Barnes and Allan [78] considered a model of
1/𝑓 noise explicitly based on fractional integration. Granger’s observation was followed up by
both himself, and independentlyin hydrology by Jonathan Hosking [79], who between them
laid the foundations for a different class of long memory model. This class of ARFIMA
models are the most commonly used long memory models of the present day. If the empirical
findings of Hurst helped to stimulate the field, and the models of Mandelbrot helped to
revolutionise the field, the class of ARFIMA models can be said to have made the field
accessible to all.

4. Fractionally Differenced Models

Hosking and Granger’s ARFIMA(p,d,q) process 𝑋𝑡 is defined (see Beran [1] and Beran et
al. [2] for details) as

𝜙(ℬ) (1 − ℬ)𝑑 𝑋𝑡 = 𝜓(ℬ) 𝜀𝑡 (2)

Here, the backshift operator is defined by ℬ𝑋𝑡 = 𝑋𝑡−1. The polynomials


𝜙 = 1 − ∑𝑝𝑘=1 𝜙𝑘 𝑧𝑘
and 𝜓 = 1 + ∑𝑞𝑘=1 𝜓𝑘 𝑧𝑘
describe the autoregressive and moving
𝜀
average terms respectively. The innovations 𝑡 are Gaussian, stationary, zero mean,
𝜎
independent and identically distributed, with variance 𝜀2 i.e., they form a white Gaussian
𝑋
noise process. 𝑡 is second-order stationary and invertible, having a Wold decomposition of

𝑋𝑡 = ∑ 𝑎𝑗 𝜀𝑡−𝑗
(/)
(3)
𝑗=0
The coefficients 𝑎𝑗 obey buscarmenú
𝑎𝑗 = (−1)𝑗 Γ(𝑗 + 1)Γ(1Γ(1− −𝑑)𝑑 − 𝑗) (4)

and allow a power series representation of the fractional differencing terms using

∞ 𝑗
𝐴(𝑧) = (1 − 𝑧) = ∑ 𝑎𝑗 𝑧
−𝑑 (5)
𝑗=0
This expansion allows the spectral density function 𝑓(.) of 𝑋𝑡 to be obtained. Its
behaviour at the origin is found to be

𝜎𝜀2 |𝜓(1)|2 |𝜆|−2𝑑


𝑓 (𝜆) ∼ 2𝜋 |𝜙(1)|2 (6)

which may be compared with the leading order behaviour of FGN at the origin:

𝑓 (𝜆) ∼ 𝑐𝑓 |𝜆|1−2𝐻 (7)

so 𝑑 = 𝐻 − 1/2 . One of the objections to Mandelbrot’s fractional Gaussian noise was that it
was a discrete approximation to a continuous process. Hosking [79] explained how FGN can
be roughly thought of as the discrete version of a fractional derivative of Brownian motion. In
other words, FGN is obtained by fractionally differentiating, then discretising. Hosking
proposed to reverse this order of operations, i.e discretising first, then fractionally
differencing.
The advantage of this approach is that the discrete version of Brownian motion has an
intuitive interpretation; it is the simple random walk, or ARIMA (0,1,0) model. We may
fractionally difference this using the well-defined ‘fractional differencing operator of order d’
to obtain the ARFIMA ( 0 , 𝑑 ,0)
proceso, que para 0 < 𝑑 < 1/2 es estacionario y posee una
larga memoria. A partir de esta derivación vaga, vemos inmediatamente una clara ventaja de
este proceso: es formalizable como una simple extensión de los modelos ARIMA clásicos de
Box-Jenkins.
Granger y Joyeux [ 80 ] llegaron a una conclusión similar al observar que era posible
diferenciar fraccionariamente un proceso y, para no sobre o subdiferenciar los datos, puede
ser deseable hacerlo. La motivación directa fue proporcionada por [ 81 ], quien demostró que
tales procesos podrían surgir como una agregación de procesos AR (1) independientes,
donde los parámetros autorregresivos se distribuyeron de acuerdo con una distribución Beta
(/)
(esta agregación de variables microeconómicas era una verdadera motivación, en lugar de un

buscarmen
ejemplo artificial). Además, Granger y Joyeux señalaron que en los pronósticos a largo plazo
el componente de baja frecuencia es el de mayor importancia. Vale la pena señalar que ú la
previsión es bastante diferente de la síntesis . discussed earlier; the former takes an observed
sequence and, based on a statistical examination of its past, attempts to extrapolate its
future. This is a deterministic approach and, given the same data and using the same
methods, two practitioners will produce the same forecasts. Synthesis on the other hand is a
method of producing a representative sample path of a given process and is therefore
stochastic in nature. Given the same model and parameters, two practitioners will produce
different sample paths (assuming their random number generator seeds are not initiated to
the same value). However, their sequences will have the same statistical properties.
Tanto Granger como Joyeux [ 80 ] y Hosking [ 79 ] reconocieron que su modelo se
basaba en supuestos subyacentes diferentes a los modelos de Mandelbrot. También
reconocieron la extrema utilidad de introducir la memoria larga en el marco de Box-Jenkins.
Al considerar su modelo fraccionariamente diferenciado como ARIMA (0, 𝑑, 0)
proceso, fue un
salto obvio incluir los parámetros 𝑝, 𝑞
para modelar efectos a corto plazo; de ahí, el ARFIMA
completo (𝑝, 𝑑, 𝑞)
modelo. Al desarrollar un proceso que podía modelar las propiedades de la
memoria corta y larga, los autores eliminaron la dicotomía forzada entre los modelos ARMA y
FGN. Al poder modelar ambos tipos de memoria simultáneamente, los modelos ARFIMA
resolvieron inmediatamente la principal objeción práctica al modelo FGN de ​Mandelbrot.
Dependiendo del contexto y punto de vista individual, los modelos ARFIMA pueden verse
como modelos puros de memoria corta ajustados para inducir un comportamiento de
memoria larga, o modelos puros de memoria larga ajustados para tener en cuenta el
comportamiento a corto plazo. Los modelos ARFIMA se introducen con mayor frecuencia
utilizando la primera de estas interpretaciones, presumiblemente porque la mayoría de los
profesionales encuentran los modelos elementales de Box-Jenkins antes de la memoria; sin
embargo, podría decirse que es más útil considerar la última interpretación.
Aunque tardaron en despegar, la mayor flexibilidad de los modelos ARFIMA y su facilidad
de uso general en comparación con el FGN de ​Mandelbrot significaron que gradualmente se
convirtieron en el modelo de memoria larga preferido en muchas áreas, incluidas la
hidrología y la econometría, aunque hemos descubierto que todavía funcionan. ser menos
conocido en física que FGN. Aparte de su discreción (que puede ser una desventaja o no
según el punto de vista), la única desventaja que tienen los modelos ARFIMA es que ya no
son completamente autosimilares. Las sumas parciales reescaladas de un ARFIMA “puro”
( 0 , 𝑑 , El0 modelo
) converge en distribución a FBM (ver, por ejemplo, [ 82 ], §6), por lo que, en
cierto sentido, el proceso puede verse como los incrementos de un proceso asintóticamente
autosemejante. Sin embargo, cualquier componente no trivial de la memoria corta ( p o q )
introduce un "tic" temporal y destruye esta autosimilitud.
Tal vez inevitablemente, dada su motivación original para introducir la autosimilitud
(/)
como explicación del fenómeno Hurst, y su posterior desarrollo de todo el concepto de

buscarmen
escalamiento en la teoría fractal, Mandelbrot no se sintió atraído por los modelos ARFIMA.
Décadas después de su introducción, y a pesar de su popularidad, Mandelbrot afirmaría: ú

[Granger] prefiere una versión de tiempo discreto de FBM que difiere un poco del
algoritmo Tipo I y Tipo II en [ 15 ]. La discretización suele estar motivada por una
conveniencia incuestionable, pero yo la veo como algo más que un detalle. Prefiero
mucho los modelos que poseen propiedades de invariancia temporal o escalamiento.
En estos modelos ningún intervalo de tiempo se privilegia por ser intrínseco. Por el
contrario, en los modelos de tiempo discreto, un intervalo de tiempo privilegiado se
impone de manera no intrínseca.

[ 83 ]

Sin embargo, la conveniencia parece ser la que manda en las estadísticas, ya que la
inferencia basada en ARFIMA se aplica en la práctica con mucha más frecuencia que
FBM/FGN. Muchos profesionales argumentarían que no es difícil justificar el uso de un
“intervalo de tiempo privilegiado” en un verdadero contexto de análisis de datos: el intervalo
en el que se muestrean los datos y/o en el que normalmente se tomarían decisiones basadas
en dichos datos, Siempre disfrutamos del privilegio de modelar e inferir.
As we saw above, the introduction of the LRD concept into science came with
Mandelbrot’s application of the fractional Brownian models of Kolmogorov to an
environmetric observation—Hurst’s effect in hydrology. Nowadays, an important new
environmetric application for LRD is to climate research. Here, ARFIMA plays an important
role in understanding long-term climate variability and in trend estimation [84] but remains
less well known in some user communities compared to, for example, SRD models of the
Box-Jenkins type, of which AR(1) is still the most frequently applied. Conversely, in many
𝛼
branches of physics the fractional -stable family of models including FBM remain rather
better known than ARFIMA. The process of codifying the reasons for the similarities and
differences between these models, and also the closely related anomalous diffusion models
such as the Continuous Time Random Walk, in a way accessible to users, is under way but
much more remains to be done here, particularly on the “physics of FARIMA”.

5. Conclusions

We have attempted to demonstrate the original motivation behind long memory


processes, and trace the early evolution of the conceptof long memory, from the early 1950s
to the late 1970s. Debates over the nature of such processes, and their applicability or
appropriateness to reality, are still ongoing.Importantly, the physical meaning of FBM has
been clarified by studies which show how it plays the role of the noise term in the generalised
Langevin equation when a particular (“
(/)
1/𝑓
”) choice of heat bath spectral density has been
made and when a fluctuation-dissipation theorem applies, see for example [85]. In the

buscarmen
mathematical, statistical and econometric communities, several mechanisms for LRD have
been investigated, including the aggregation referred to earlier, and some which emulate ú
LRD
behaviour by regime switching [86] or trends (see also [71]). Further discussion of these
𝑅/𝑆
topics can be found in [10]. The initial diagnostic of LRD has been further developed, e.g.,
by Lo [87], and there is now a very extensive mathematical and statistical literature on
estimation and testing of long memory using both parametric and non-parametric methods,
reviewed for example in Chapter 5 of [2].
Rather than draw our own conclusions, we intended to illuminate the story of this
fascinating area of science, and in particular the role played by Benoit Mandelbrot, who died
in 2010. The facet of Mandelbrot’s genius on show here was to use his strongly geometrical
mathematical imagination to link some very arcane aspects of the theory of stochastic
processes to the needs of operational environmetric statistics. Quite how remarkable this
was can only be fully appreciated when one reminds oneself of the available data and
computational resources of the early 1960s, even at IBM. The wider story [6,7] in which this
𝛼
paper’s theme is embedded, of how he developed and applied in sequence, first the -stable
model in economics, followed by the fractional renewal model in 1/𝑓noise, and then FBM,
and a fractional hyperbolic precursor to the linear fractional stable models, and finally a
multifractal model, all in the space of about 10 years, shows both mathematical creativity and
a real willingness to listen to what the data was telling him. The fact the he (and his critics)
were perhaps less willing to listen to each other is a human trait whose effects on this story—
we trust—will become less significant over time.

Acknowledgments

Open access publication was supported by funds from the RCUK at the University of
Warwick. RBG acknowledges EPSRC funding at Cambridge Statslab under EPSRC:
EP/D065704/1. CF was supported by the German Science Foundation (DFG) through the
cluster of excellence CliSAP and the collaborative research center TRR181, while NWW has
recently been supported by ONR NICOP grant N62909-15-1-N143 to Warwick. He also
acknowledges support from the London Mathematical Laboratory and travel support from
KLIMAFORSK project number 229754. NWW and CF acknowledge the stimulating
environment of the British Antarctic Survey during the initial development of this paper. We
thank Cosma Shalizi, Holger Kantz, and the participants in the 2013–4 International Space
Science Institute programme on “Self-Organized Criticality and Turbulence” for discussions,
and David Spiegelhalter for his help.
Author Contributions
(/)
This paper is derived from Appendix D of TG’s PhD thesis [88]. All the authors were

manuscript.
buscarmen
involved in the preparation of the manuscript. All the authors have read and approved the final
ú

Conflicts of Interest

The authors declare no conflict of interest.

Abbreviations

ACF Auto-Correlation Function

ARFIMA AutoRegressive Fractionally Integrated Moving Average

BDoA Brownian Domain of Attraction

FBM Fractional Brownian Motion

FGN Fractional Gaussian Noise

LRD Long Range Dependence

SRD Short Range Dependence

References

1. Beran, J. Statistics for Long Memory Processes; Chapman & Hall: New York, NY, USA,
1994. [Google Scholar (https://scholar.google.com/scholar_lookup?
title=Statistics+for+Long+Memory+Processes&author=Beran,+J.&publication_year=1
994)]
2. Beran, J.; Feng, Y.; Ghosh, S.; Kulik, R. Long Memory Processes; Springer: Heidelberg,
Germany, 2013. [Google Scholar (https://scholar.google.com/scholar_lookup?
title=Long+Memory+Processes&author=Beran,+J.&author=Feng,+Y.&author=Ghosh,+
S.&author=Kulik,+R.&publication_year=2013)]
3. Sornette, D. Why Stock Markets Crash: Critical Events in Complex Financial Systems;
Princeton University Press: Princeton, NJ, USA, 2004. [Google Scholar
(https://scholar.google.com/scholar_lookup?
title=Why+Stock+Markets+Crash:+Critical+Events+in+Complex+Financial+Systems&a
uthor=Sornette,+D.&publication_year=2004)]
4. Franzke,
(/)
C. Nonlinear trends, long-range dependence and climate noise properties of
surface air temperature. J. Clim. 2012, 25, 4172–4183. [Google Scholar

buscarmenú
(https://scholar.google.com/scholar_lookup?title=Nonlinear+trends,+long-
range+dependence+and+climate+noise+properties+of+surface+air+temperature&aut
hor=Franzke,+C.&publication_year=2012&journal=J.+Clim.&volume=25&pages=4172
%E2%80%934183&doi=10.1175/JCLI-D-11-00293.1)] [CrossRef
(https://doi.org/10.1175/JCLI-D-11-00293.1)]
5. Montanari, A. Long-Range Dependence in Hydrology. In Theory and Applications of
Long-Range Dependence; Doukhan, P., Oppenheim, G., Taqqu, M.S., Eds.; Birkhäuser:
Boston, MA, USA, 2003; pp. 461–472. [Google Scholar
(https://scholar.google.com/scholar_lookup?title=Long-
Range+Dependence+in+Hydrology&author=Montanari,+A.&publication_year=2003&p
ages=461%E2%80%93472)]
6. Mandelbrot, B.B.; Hudson, R.L. The (mis)Behaviour of Markets: A Fractal View of Risk,
Ruin, and Reward, 2nd ed.; Profile Books: London, UK, 2008. [Google Scholar
(https://scholar.google.com/scholar_lookup?title=The+
(mis)Behaviour+of+Markets:+A+Fractal+View+of+Risk,+Ruin,+and+Reward&author=
Mandelbrot,+B.B.&author=Hudson,+R.L.&publication_year=2008)]
7. Mandelbrot, B.B. The Fractalist: Memoir of a Scientific Maverick; Vintage Books: New
York, NY, USA, 2013. [Google Scholar (https://scholar.google.com/scholar_lookup?
title=The+Fractalist:+Memoir+of+a+Scientific+Maverick&author=Mandelbrot,+B.B.&p
ublication_year=2013)]
8. Taqqu, M.S. Benoit Mandelbrot and Fractional Brownian Motion. Stat. Sci. 2013, 28,
131–134. [Google Scholar (https://scholar.google.com/scholar_lookup?
title=Benoit+Mandelbrot+and+Fractional+Brownian+Motion&author=Taqqu,+M.S.&pu
blication_year=2013&journal=Stat.+Sci.&volume=28&pages=131%E2%80%93134&do
i=10.1214/12-STS389)] [CrossRef (https://doi.org/10.1214/12-STS389)]
9. Brockwell, P.J.; Davis, R.A. Time Series: Theory and Methods, 2nd ed.; Springer: New
York, NY, USA, 1991. [Google Scholar (https://scholar.google.com/scholar_lookup?
title=Time+Series:+Theory+and+Methods&author=Brockwell,+P.J.&author=Davis,+R.
A.&publication_year=1991)]
10. Samorodnitsky, G. Long Range Dependence. Found. Trends Stoch. Syst. 2006, 1, 163–
257. [Google Scholar (https://scholar.google.com/scholar_lookup?
title=Long+Range+Dependence&author=Samorodnitsky,+G.&publication_year=2006&
journal=Found.+Trends+Stoch.+Syst.&volume=1&pages=163%E2%80%93257&doi=1
0.1561/0900000004)] [CrossRef (https://doi.org/10.1561/0900000004)]
11. Baillie,
(/)
R.T. Long memory processes and fractional integration in econometrics. J.
Econom. 1996, 73, 5–59. [Google Scholar
(https://scholar.google.com/scholar_lookup?
buscarmenú
title=Long+memory+processes+and+fractional+integration+in+econometrics&author
=Baillie,+R.T.&publication_year=1996&journal=J.+Econom.&volume=73&pages=5%E
2%80%9359&doi=10.1016/0304-4076(95)01732-1)] [CrossRef
(https://doi.org/10.1016/0304-4076(95)01732-1)]
12. Aharony, A.; Feder, J. Fractals in Physics: Essays in Honour of Benoit B. Mandelbrot;
North-Holland: Amsterdam, The Netherlands; New York, NY, USA, 1990. [Google Scho‐
lar (https://scholar.google.com/scholar_lookup?
title=Fractals+in+Physics:+Essays+in+Honour+of+Benoit+B.+Mandelbrot&author=Ah
arony,+A.&author=Feder,+J.&publication_year=1990)]

13. Turcotte, D. Fractals and Chaos in Geology and Geophysics, 2nd ed.; Cambridge
University Press: Cambridge, UK, 1997. [Google Scholar
(https://scholar.google.com/scholar_lookup?
title=Fractals+and+Chaos+in+Geology+and+Geophysics&author=Turcotte,+D.&public
ation_year=1997)]
14. Hurst, H.E. Long-term storage capacity of reservoirs. Trans. Am. Soc. Civ. Eng. 1951,
116, 770–799, With dicussion. [Google Scholar
(https://scholar.google.com/scholar_lookup?title=Long-
term+storage+capacity+of+reservoirs&author=Hurst,+H.E.&publication_year=1951&j
ournal=Trans.+Am.+Soc.+Civ.+Eng.&volume=116&pages=770%E2%80%93799)]
15. Mandelbrot, B.B.; Wallis, J.R. Computer experiments with Fractional Gaussian noises.
Water Resour. Res. 1969, 5, 228–267. [Google Scholar
(https://scholar.google.com/scholar_lookup?
title=Computer+experiments+with+Fractional+Gaussian+noises&author=Mandelbrot,
+B.B.&author=Wallis,+J.R.&publication_year=1969&journal=Water+Resour.+Res.&vol
ume=5&pages=228%E2%80%93267&doi=10.1029/WR005i001p00228)] [CrossRef
(https://doi.org/10.1029/WR005i001p00228)]

16. Rippl, W. The Capacity of storage-reservoirs for water-supply. Min. Proc. Inst. Civ. Eng.
1883, 71, 270–278. [Google Scholar (https://scholar.google.com/scholar_lookup?
title=The+Capacity+of+storage-reservoirs+for+water-
supply&author=Rippl,+W.&publication_year=1883&journal=Min.+Proc.+Inst.+Civ.+Eng
.&volume=71&pages=270%E2%80%93278)]
17. Klemeš,
(/)
V. One hundred years of applied storage reservoir theory. Water Resour.
Manag. 1987, 1, 159–175. [Google Scholar
(https://scholar.google.com/scholar_lookup?
buscarmenú
title=One+hundred+years+of+applied+storage+reservoir+theory&author=Kleme%C5%
A1,+V.&publication_year=1987&journal=Water+Resour.+Manag.&volume=1&pages=1
59%E2%80%93175&doi=10.1007/BF00429941)] [CrossRef
(https://doi.org/10.1007/BF00429941)]
18. Hazen, A. Storage to be provided in impounding reservoirs for municipal water supply.
Proc. Am. Soc. Civ. Eng. 1913, 39, 1943–2044, With dicussion. [Google Scholar
(https://scholar.google.com/scholar_lookup?
title=Storage+to+be+provided+in+impounding+reservoirs+for+municipal+water+supp
ly&author=Hazen,+A.&publication_year=1913&journal=Proc.+Am.+Soc.+Civ.+Eng.&vo
lume=39&pages=1943%E2%80%932044)]
19. Sudler, C. Storage required for the regulation of streamflow. Trans. Am. Soc. Civ. Eng.
1927, 91, 622–660. [Google Scholar (https://scholar.google.com/scholar_lookup?
title=Storage+required+for+the+regulation+of+streamflow&author=Sudler,+C.&public
ation_year=1927&journal=Trans.+Am.+Soc.+Civ.+Eng.&volume=91&pages=622%E2%
80%93660)]
20. Moran, P.A.P. The Theory of Storage; Wiley: New York, NY, USA, 1959. [Google Scholar
(https://scholar.google.com/scholar_lookup?
title=The+Theory+of+Storage&author=Moran,+P.A.P.&publication_year=1959)]
21. Lloyd, E.H. Stochastic reservoir theory. Adv. Hydrosci. 1967, 4, 281. [Google Scholar
(https://scholar.google.com/scholar_lookup?
title=Stochastic+reservoir+theory&author=Lloyd,+E.H.&publication_year=1967&journ
al=Adv.+Hydrosci.&volume=4&pages=281)]
22. Feller, W. The Asymptotic Distribution of the Range of Sums of Independent Random
Variables. Ann. Math. Stat. 1951, 22, 427–432. [Google Scholar
(https://scholar.google.com/scholar_lookup?
title=The+Asymptotic+Distribution+of+the+Range+of+Sums+of+Independent+Rando
m+Variables&author=Feller,+W.&publication_year=1951&journal=Ann.+Math.+Stat.&v
olume=22&pages=427%E2%80%93432&doi=10.1214/aoms/1177729589)] [CrossRef
(https://doi.org/10.1214/aoms/1177729589)]
23. Chow,
(/)
V.T. Discussion of “Long-term storage capacity of reservoirs”. Trans. Am. Soc.
Civ. Eng. 1951, 116, 800–802. [Google Scholar

buscarmenú
(https://scholar.google.com/scholar_lookup?title=Discussion+of+%E2%80%9CLong-
term+storage+capacity+of+reservoirs%E2%80%9D&author=Chow,+V.T.&publication_y
ear=1951&journal=Trans.+Am.+Soc.+Civ.+Eng.&volume=116&pages=800%E2%80%9
3802)]

24. Hurst, H.E. Methods of Using Long-term storage in reservoirs. Proc. Inst. Civ. Eng. 1956,
5, 519–590, With dicussion. [Google Scholar
(https://scholar.google.com/scholar_lookup?title=Methods+of+Using+Long-
term+storage+in+reservoirs&author=Hurst,+H.E.&publication_year=1956&journal=Pr
oc.+Inst.+Civ.+Eng.&volume=5&pages=519%E2%80%93590&doi=10.1680/iicep.1956
.11503)] [CrossRef (https://doi.org/10.1680/iicep.1956.11503)]

25. Hurst, H.E. The Problem of Long-Term Storage in Reservoirs. Hydrol. Sci. J. 1956, 1,
13–27. [Google Scholar (https://scholar.google.com/scholar_lookup?
title=The+Problem+of+Long-
Term+Storage+in+Reservoirs&author=Hurst,+H.E.&publication_year=1956&journal=H
ydrol.+Sci.+J.&volume=1&pages=13%E2%80%9327&doi=10.1080/026266656094936
44)] [CrossRef (https://doi.org/10.1080/02626665609493644)]
26. Hurst, H.E.; Black, R.P.; Simaika, Y.M. Long-Term Storage: An Experimental Study;
Constable: London, UK, 1965. [Google Scholar
(https://scholar.google.com/scholar_lookup?title=Long-
Term+Storage:+An+Experimental+Study&author=Hurst,+H.E.&author=Black,+R.P.&au
thor=Simaika,+Y.M.&publication_year=1965)]

27. Anis, A.A.; Lloyd, E.H. On the range of partial sums of a finite number of independent
normal variates. Biometrika 1953, 40, 35–42. [Google Scholar
(https://scholar.google.com/scholar_lookup?
title=On+the+range+of+partial+sums+of+a+finite+number+of+independent+normal+v
ariates&author=Anis,+A.A.&author=Lloyd,+E.H.&publication_year=1953&journal=Bio
metrika&volume=40&pages=35%E2%80%9342&doi=10.1093/biomet/40.1-2.35)]
[CrossRef (https://doi.org/10.1093/biomet/40.1-2.35)]
28. Solari,
(/)
YO; Anis, AA La media y la varianza del máximo de las sumas parciales
ajustadas de un número finito de variables normales independientes. Ana.

buscarmenú
Matemáticas. Estadística. 1957 , 28 , 706–716. [ Google Scholar
(https://scholar.google.com/scholar_lookup?
title=The+Mean+and+Variance+of+the+Maximum+of+the+Adjusted+Partial+Sums+of
+a+Finite+Number+of+Independent+Normal+Variates&author=Solari,+M.E.&author=A
nis,+A.A.&publication_year=1957&journal=Ann.+Math.+Stat.&volume=28&pages=70
6%E2%80%93716&doi=10.1214/aoms/1177706882) ] [ CrossRef
(https://doi.org/10.1214/aoms/1177706882) ]
29. Moran, PAP Sobre el rango de sumas acumuladas. Ana. Inst. Estadística. Matemáticas.
1964 , 16 , 109-112. [ Google Scholar (https://scholar.google.com/scholar_lookup?
title=On+the+range+of+cumulative+sums&author=Moran,+P.A.P.&publication_year=1
964&journal=Ann.+Inst.+Stat.+Math.&volume=16&pages=109%E2%80%93112&doi=1
0.1007/BF02868565) ] [ CrossRef (https://doi.org/10.1007/BF02868565) ]
30. Mandelbrot, BB; Wallis, JR Robustez del rango reescalado R / S en la medición de la
dependencia estadística no cíclica de largo plazo. Recurso Acuático. Res. 1969 , 5 ,
967–988. [ Google Scholar (https://scholar.google.com/scholar_lookup?
title=Robustness+of+the+rescaled+range+R/S+in+the+measurement+of+noncyclic+l
ong+run+statistical+dependence&author=Mandelbrot,+B.B.&author=Wallis,+J.R.&pub
lication_year=1969&journal=Water+Resour.+Res.&volume=5&pages=967%E2%80%93
988&doi=10.1029/WR005i005p00967) ] [ CrossRef
(https://doi.org/10.1029/WR005i005p00967) ]
31. Mandelbrot, BB Teoremas de límite en el rango del puente autonormalizado. Zeitschrift
für Wahrscheinlichkeitstheorie und Verwandte Gebiete 1975 , 31 , 271–285. [ Google
Scholar (https://scholar.google.com/scholar_lookup?
title=Limit+theorems+on+the+self-
normalized+bridge+range&author=Mandelbrot,+B.B.&publication_year=1975&journal
=Zeitschrift+f%C3%BCr+Wahrscheinlichkeitstheorie+und+Verwandte+Gebiete&volum
e=31&pages=271%E2%80%93285&doi=10.1007/BF00532867) ] [ CrossRef
(https://doi.org/10.1007/BF00532867) ]
32. Mandelbrot, BB; Taqqu, MS Análisis robusto R / S de correlación serial a largo plazo. En
Boletín del Instituto Internacional de Estadística, Actas del 42.º período de sesiones del
Instituto Internacional de Estadística, Manila, Filipinas, 4 a 14 de diciembre de 1979 ;
Instituto Internacional de Estadística (ISI): La Haya, Países Bajos, 1979; Volumen 48,
págs. 69-105, con discusión. [ Google Académico
(https://scholar.google.com/scholar_lookup?title=Robust+R/S+analysis+of+long-
run+serial+correlation&author=Mandelbrot,+B.B.&author=Taqqu,+M.S.&publication_y
ear=1979&pages=69%E2%80%93105) ]
33. O'Connell,
(/)
PE Un modelo estocástico simple de la ley de Hurst. En Mathematical
Models in Hydrology, Actas del Simposio de Varsovia, Varsovia, Polonia, julio de 1971 ;

buscarmenú
Asociación Internacional de Ciencias Hidrológicas: París, Francia, 1974; Volumen 1,
págs. 169–187. [ Google Académico (https://scholar.google.com/scholar_lookup?
title=A+simple+stochastic+modelling+of+Hurst%E2%80%99s+law&author=O%E2%80
%99Connell,+P.E.&publication_year=1974&pages=169%E2%80%93187) ]

34. Hurst, HE Un modelo estadístico sugerido de algunas series de tiempo que ocurren en
la naturaleza. Naturaleza 1957 , 180 , 494. [ Google Scholar
(https://scholar.google.com/scholar_lookup?
title=A+suggested+statistical+model+of+some+time+series+which+occur+in+nature
&author=Hurst,+H.E.&publication_year=1957&journal=Nature&volume=180&pages=4
94&doi=10.1038/180494a0) ] [ CrossRef (https://doi.org/10.1038/180494a0) ]

35. Klemeš, V. El fenómeno Hurst: ¿un rompecabezas? Recurso Acuático. Res. 1974 , 10 ,
675–688. [ Google Scholar (https://scholar.google.com/scholar_lookup?
title=The+Hurst+Phenomenon:+A+puzzle?
&author=Kleme%C5%A1,+V.&publication_year=1974&journal=Water+Resour.+Res.&v
olume=10&pages=675%E2%80%93688&doi=10.1029/WR010i004p00675) ] [ Cross‐
Ref (https://doi.org/10.1029/WR010i004p00675) ]
36. Potter, KW Evidencia de la no estacionariedad como explicación física del fenómeno
de Hurst. Recurso Acuático. Res. 1976 , 12 , 1047–1052. [ Google Scholar
(https://scholar.google.com/scholar_lookup?
title=Evidence+for+nonstationarity+as+a+physical+explanation+of+the+Hurst+Pheno
menon&author=Potter,+K.W.&publication_year=1976&journal=Water+Resour.+Res.&v
olume=12&pages=1047%E2%80%931052&doi=10.1029/WR012i005p01047) ] [
CrossRef (https://doi.org/10.1029/WR012i005p01047) ]
37. Bhattacharya, enfermera registrada; Gupta, VK; Waymire, E. El efecto Hurst según las
tendencias. J. Aplica. Probablemente. 1983 , 20 , 649–662. [ Google Scholar
(https://scholar.google.com/scholar_lookup?
title=The+Hurst+Effect+under+Trends&author=Bhattacharya,+R.N.&author=Gupta,+V.
K.&author=Waymire,+E.&publication_year=1983&journal=J.+Appl.+Probab.&volume=
20&pages=649%E2%80%93662&doi=10.1017/S0021900200023895) ] [ CrossRef
(https://doi.org/10.1017/S0021900200023895) ]
38. Barnard, GA Discusión sobre “Métodos de uso del almacenamiento a largo plazo en
embalses”. Proc. Inst. Civilización. Ing. 1956 , 5 , 552–553. [ Google Académico
(https://scholar.google.com/scholar_lookup?
title=Discussion+of+%E2%80%9CMethods+of+Using+Long-
term+storage+in+reservoirs%E2%80%9D&author=Barnard,+G.A.&publication_year=19
56&journal=Proc.+Inst.+Civ.+Eng.&volume=5&pages=552%E2%80%93553) ]
39. Mandelbrot,
(/)
BB; Wallis, JR Noah, Joseph y la hidrología operativa. Recurso Acuático.
Res. 1968 , 4 , 909–918. [ Google Scholar
(https://scholar.google.com/scholar_lookup?
buscarmenú
title=Noah,+Joseph+and+operational+hydrology&author=Mandelbrot,+B.B.&author=
Wallis,+J.R.&publication_year=1968&journal=Water+Resour.+Res.&volume=4&pages
=909%E2%80%93918&doi=10.1029/WR004i005p00909) ] [ CrossRef
(https://doi.org/10.1029/WR004i005p00909) ]
40. Siddiqui, MM La distribución asintótica del rango y otras funciones de sumas parciales
de procesos estacionarios. Recurso Acuático. Res. 1976 , 12 , 1271–1276. [ Google
Scholar (https://scholar.google.com/scholar_lookup?
title=The+asymptotic+distribution+of+the+range+and+other+functions+of+partial+su
ms+of+stationary+processes&author=Siddiqui,+M.M.&publication_year=1976&journa
l=Water+Resour.+Res.&volume=12&pages=1271%E2%80%931276&doi=10.1029/WR
012i006p01271) ] [ CrossRef (https://doi.org/10.1029/WR012i006p01271) ]
41. Matalas, Carolina del Norte; Huzzen, CS Una propiedad del rango de sumas parciales.
En Actas del Simposio Internacional de Hidrología; Universidad Estatal de Colorado:
Fort Collins, CO, EE.UU., 1967; Volumen 1, págs. 252–257. [ Google Académico
(https://scholar.google.com/scholar_lookup?
title=A+property+of+the+range+of+partial+sums&conference=Proceedings+of+the+I
nternational+Hydrology+Symposium&author=Matalas,+N.C.&author=Huzzen,+C.S.&p
ublication_year=1967&pages=252%E2%80%93257) ]
42. Fiering, Síntesis MB Streamflow ; Harvard University Press: Cambridge MA, EE.UU.,
1967. [ Google Scholar (https://scholar.google.com/scholar_lookup?
title=Streamflow+Synthesis&author=Fiering,+M.B.&publication_year=1967) ]
43. Bachelier, L. Théorie de la Spéculation. Annales Scientifiques de l'École Normale
Supérieure 1900 , 3 , 21–86. [ Google Scholar
(https://scholar.google.com/scholar_lookup?
title=Th%C3%A9orie+de+la+Sp%C3%A9culation&author=Bachelier,+L.&publication_y
ear=1900&journal=Annales+Scientifiques+de+l%E2%80%99%C3%89cole+Normale+S
up%C3%A9rieure&volume=3&pages=21%E2%80%9386&doi=10.24033/asens.476) ] [
CrossRef (https://doi.org/10.24033/asens.476) ]
44. Mandelbrot, BB La variación de ciertos precios especulativos. J. Autobús. 1963 , 36 ,
394–419. [ Google Scholar (https://scholar.google.com/scholar_lookup?
title=The+Variation+of+Certain+Speculative+Prices&author=Mandelbrot,+B.B.&public
ation_year=1963&journal=J.+Bus.&volume=36&pages=394%E2%80%93419&doi=10.
1086/294632) ] [ CrossRef (https://doi.org/10.1086/294632) ]
45. Samorodnitsky,
(/)
G.; Taqqu, MS Procesos aleatorios estables no gaussianos: modelos
estocásticos con varianza infinita ; Chapman & Hall: Nueva York, NY, EE.UU., 1994. [
Google Scholar
buscarmenú
(https://scholar.google.com/scholar_lookup?title=Stable+Non-
Gaussian+Random+Processes:+Stochastic+Models+with+Infinite+Variance&author=
Samorodnitsky,+G.&author=Taqqu,+M.S.&publication_year=1994) ]
46. Cootner, PH comenta sobre “La variación de ciertos precios especulativos”. En El
carácter aleatorio de los precios del mercado de valores ; Prensa del MIT: Cambridge,
MA, EE.UU., 1964; págs. 333–337. [ Google Académico
(https://scholar.google.com/scholar_lookup?
title=Comments+on+%E2%80%9CThe+Variation+of+Certain+Speculative+Prices%E2
%80%9D&author=Cootner,+P.H.&publication_year=1964&pages=333%E2%80%93337)
]

47. Mandelbrot, BB Las leyes de potencia experimentales sugieren que el escalamiento


autoafín es ubicuo en la naturaleza. En Autoafinidad gaussiana y fractales: globalidad, la
Tierra, ruido 1/f y R/S ; Springer-Verlag: Nueva York, NY, EE.UU., 2002; págs. 187-203. [
Google Académico (https://scholar.google.com/scholar_lookup?
title=Experimental+power-laws+suggest+that+self-
affine+scaling+is+ubiquitous+in+nature&author=Mandelbrot,+B.B.&publication_year=
2002&pages=187%E2%80%93203) ]
48. Mandelbrot, BB Une classe de processus stochastiques homothétiques à soi;
Aplicación a la ley climatológica de HE Hurst. Comptes Rendus (París) 1965 , 260 ,
3274–3277. [ Google Académico (https://scholar.google.com/scholar_lookup?
title=Une+classe+de+processus+stochastiques+homoth%C3%A9tiques+%C3%A0+soi
;+application+%C3%A0+loi+climatologique+de+H.+E.+Hurst&author=Mandelbrot,+B.B
.&publication_year=1965&journal=Comptes+Rendus+
(Paris)&volume=260&pages=3274%E2%80%933277) ]
49. Mandelbrot, BB; Van Ness, JW Movimientos brownianos fraccionales, ruidos
fraccionarios y aplicaciones. SIAM Rev. 1968 , 10 , 422–437. [ Google Scholar
(https://scholar.google.com/scholar_lookup?
title=Fractional+Brownian+Motions,+Fractional+Noises+and+Applications&author=M
andelbrot,+B.B.&author=Van+Ness,+J.W.&publication_year=1968&journal=SIAM+Rev
.&volume=10&pages=422%E2%80%93437&doi=10.1137/1010093) ] [ CrossRef
(https://doi.org/10.1137/1010093) ]
50. Adelman, I. Ciclo largo: ¿hecho o artefacto? Soy. Economía. Rev. 1965 , 55 , 444–463. [
Google Académico (https://scholar.google.com/scholar_lookup?
title=Long+Cycle%E2%80%94Fact+or+Artifact?
&author=Adelman,+I.&publication_year=1965&journal=Am.+Econ.+Rev.&volume=55&
pages=444%E2%80%93463) ]
51. Granger,
(/)
CWJ La forma espectral típica de una variable económica. Econométrica 1966
, 34 , 150-161. [ Google Scholar (https://scholar.google.com/scholar_lookup?

buscarmenú
title=The+Typical+Spectral+Shape+of+an+Economic+Variable&author=Granger,+C.W.
J.&publication_year=1966&journal=Econometrica&volume=34&pages=150%E2%80%
93161&doi=10.2307/1909859) ] [ CrossRef (https://doi.org/10.2307/1909859) ]
52. Mandelbrot, BB Algunos ruidos con espectro 1/ f , un puente entre la corriente
continua y el ruido blanco. Traducción IEEE. inf. Teoría 1967 , 13 , 289–298. [ Google
Scholar (https://scholar.google.com/scholar_lookup?
title=Some+noises+with+1/f+spectrum,+a+bridge+between+direct+current+and+whit
e+noise&author=Mandelbrot,+B.B.&publication_year=1967&journal=IEEE+Trans.+Inf.
+Theory&volume=13&pages=289%E2%80%93298&doi=10.1109/TIT.1967.1053992) ]
[ CrossRef (https://doi.org/10.1109/TIT.1967.1053992) ]

53. Kolmogorov, AN Wienersche Spiralen und einige andere interessante Kurven in


Hilbertschen Raum. Comptes Rendus (Doklady) 1940 , 26 , 115-118. (En alemán) [ Goo‐
gle Scholar (https://scholar.google.com/scholar_lookup?
title=Wienersche+Spiralen+und+einige+andere+interessante+Kurven+in+Hilbertsche
n+Raum&author=Kolmogorov,+A.N.&publication_year=1940&journal=Comptes+Rend
us+(Doklady)&volume=26&pages=115%E2%80%93118) ]
54. Mandelbrot, BB Un generador rápido de ruido gaussiano fraccional. Recurso Acuático.
Res. 1971 , 7 , 543–553. [ Google Scholar
(https://scholar.google.com/scholar_lookup?
title=A+fast+fractional+Gaussian+noise+generator&author=Mandelbrot,+B.B.&public
ation_year=1971&journal=Water+Resour.+Res.&volume=7&pages=543%E2%80%935
53&doi=10.1029/WR007i003p00543) ] [ CrossRef
(https://doi.org/10.1029/WR007i003p00543) ]
55. Hipel, KW; McLeod, AI Preservación del rango ajustado reescalado: 2. Estudios de
simulación utilizando modelos Box-Jenkins. Recurso Acuático. Res. 1978 , 14 , 509–
516. [ Google Scholar (https://scholar.google.com/scholar_lookup?
title=Preservation+of+the+rescaled+adjusted+range:+2.+Simulation+studies+using+
Box%E2%80%93Jenkins+Models&author=Hipel,+K.W.&author=McLeod,+A.I.&publicat
ion_year=1978&journal=Water+Resour.+Res.&volume=14&pages=509%E2%80%9351
6&doi=10.1029/WR014i003p00509) ] [ CrossRef
(https://doi.org/10.1029/WR014i003p00509) ]
56. Hipel,
(/)
KW; McLeod, AI Preservación del rango ajustado reescalado: 3. Algoritmos de
ruido gaussiano fraccional. Recurso Acuático. Res. 1978 , 14 , 517–518. [ Google Scho‐
lar
buscarmenú
(https://scholar.google.com/scholar_lookup?
title=Preservation+of+the+rescaled+adjusted+range:+3.+Fractional+Gaussian+noise+
algorithms&author=Hipel,+K.W.&author=McLeod,+A.I.&publication_year=1978&journ
al=Water+Resour.+Res.&volume=14&pages=517%E2%80%93518&doi=10.1029/WR0
14i003p00517) ] [ CrossRef (https://doi.org/10.1029/WR014i003p00517) ]
57. Davies, RB; Harte, DS Pruebas del efecto Hurst. Biometrika 1987 , 74 , 95-102. [ Google
Scholar (https://scholar.google.com/scholar_lookup?
title=Tests+for+Hurst+Effect&author=Davies,+R.B.&author=Harte,+D.S.&publication_
year=1987&journal=Biometrika&volume=74&pages=95%E2%80%93102&doi=10.109
3/biomet/74.1.95) ] [ CrossRef (https://doi.org/10.1093/biomet/74.1.95) ]

58. Taqqu, MS Nota sobre evaluaciones de R / S para ruidos fraccionarios y registros


geofísicos. Recurso Acuático. Res. 1970 , 6 , 349–350. [ Google Scholar
(https://scholar.google.com/scholar_lookup?
title=Note+on+evaluations+of+R/S+for+fractional+noises+and+geophysical+records&
author=Taqqu,+M.S.&publication_year=1970&journal=Water+Resour.+Res.&volume=
6&pages=349%E2%80%93350&doi=10.1029/WR006i001p00349) ] [ CrossRef
(https://doi.org/10.1029/WR006i001p00349) ]
59. Mandelbrot, BB; Wallis, JR Algunas propiedades a largo plazo de los registros
geofísicos. Recurso Acuático. Res. 1969 , 5 , 321–340. [ Google Scholar
(https://scholar.google.com/scholar_lookup?title=Some+long-
run+properties+of+geophysical+records&author=Mandelbrot,+B.B.&author=Wallis,+J
.R.&publication_year=1969&journal=Water+Resour.+Res.&volume=5&pages=321%E2
%80%93340&doi=10.1029/WR005i002p00321) ] [ CrossRef
(https://doi.org/10.1029/WR005i002p00321) ]
60. Lawrence, AJ; Kottegoda, NT Modelado estocástico de series temporales de caudales
fluviales. Estadística JR. Soc. Ser. A 1977 , 140 , 1–47, con discusión. [ Google Scholar
(https://scholar.google.com/scholar_lookup?
title=Stochastic+Modelling+of+Riverflow+Time+Series&author=Lawrance,+A.J.&auth
or=Kottegoda,+N.T.&publication_year=1977&journal=J.+R.+Stat.+Soc.+Ser.+A&volum
e=140&pages=1%E2%80%9347&doi=10.2307/2344516) ] [ CrossRef
(https://doi.org/10.2307/2344516) ]
61. Box, G.E.P.; Jenkins, G.M. Time Series Analysis, Forecasting and Control; Holden-Day:
San Francisco, CA, USA, 1970. [Google Scholar
(https://scholar.google.com/scholar_lookup?
title=Time+Series+Analysis,+Forecasting+and+Control&author=Box,+G.E.P.&author=J
enkins,+G.M.&publication_year=1970)]
62. Rodríguez-Iturbe,
(/)
I.; Mejia, J.M.; Dawdy, D.R. Streamflow Simulation 1: A New Look at
Markovian Models, Fractional Gaussian Noise, and Crossing Theory. Water Resour. Res.

buscarmenú
1972, 8, 921–930. [Google Scholar (https://scholar.google.com/scholar_lookup?
title=Streamflow+Simulation+1:+A+New+Look+at+Markovian+Models,+Fractional+Ga
ussian+Noise,+and+Crossing+Theory&author=Rodr%C3%ADguez-
Iturbe,+I.&author=Mejia,+J.M.&author=Dawdy,+D.R.&publication_year=1972&journal=
Water+Resour.+Res.&volume=8&pages=921%E2%80%93930&doi=10.1029/WR008i0
04p00921)] [CrossRef (https://doi.org/10.1029/WR008i004p00921)]
63. Garcia, L.E.; Dawdy, D.R.; Mejia, J.M. Long Memory Monthly Streamflow Simulation by a
Broken Line Model. Water Resour. Res. 1972, 8, 1100–1105. [Google Scholar
(https://scholar.google.com/scholar_lookup?
title=Long+Memory+Monthly+Streamflow+Simulation+by+a+Broken+Line+Model&aut
hor=Garcia,+L.E.&author=Dawdy,+D.R.&author=Mejia,+J.M.&publication_year=1972&j
ournal=Water+Resour.+Res.&volume=8&pages=1100%E2%80%931105&doi=10.1029
/WR008i004p01100)] [CrossRef (https://doi.org/10.1029/WR008i004p01100)]
64. Mejia, J.M.; Rodríguez-Iturbe, I.; Dawdy, D.R. Streamflow Simulation 2: The Broken Line
Process as a Potential Model for Hydrologic Simulation. Water Resour. Res. 1972, 8,
931–941. [Google Scholar (https://scholar.google.com/scholar_lookup?
title=Streamflow+Simulation+2:+The+Broken+Line+Process+as+a+Potential+Model+f
or+Hydrologic+Simulation&author=Mejia,+J.M.&author=Rodr%C3%ADguez-
Iturbe,+I.&author=Dawdy,+D.R.&publication_year=1972&journal=Water+Resour.+Res.
&volume=8&pages=931%E2%80%93941&doi=10.1029/WR008i004p00931)] [Cross‐
Ref (https://doi.org/10.1029/WR008i004p00931)]
65. Mejia, J.M.; Dawdy, D.R.; Nordin, C.F. Streamflow Simulation 3: The Broken Line
Process and Operational Hydrology. Water Resour. Res. 1974, 10, 242–245. [Google
Scholar (https://scholar.google.com/scholar_lookup?
title=Streamflow+Simulation+3:+The+Broken+Line+Process+and+Operational+Hydrol
ogy&author=Mejia,+J.M.&author=Dawdy,+D.R.&author=Nordin,+C.F.&publication_year
=1974&journal=Water+Resour.+Res.&volume=10&pages=242%E2%80%93245&doi=1
0.1029/WR010i002p00242)] [CrossRef (https://doi.org/10.1029/WR010i002p00242)]
66. Mandelbrot, B.B. Broken line process derived as an approximation to fractional noise.
Water Resour. Res. 1972, 8, 1354–1356. [Google Scholar
(https://scholar.google.com/scholar_lookup?
title=Broken+line+process+derived+as+an+approximation+to+fractional+noise&autho
r=Mandelbrot,+B.B.&publication_year=1972&journal=Water+Resour.+Res.&volume=8
&pages=1354%E2%80%931356&doi=10.1029/WR008i005p01354)] [CrossRef
(https://doi.org/10.1029/WR008i005p01354)]
67. Wallis,
(/)
J.R.; O’Connell, P.E. Firm Reservoir Yield—How Reliable are Historic Hydrological
Records? Hydrol. Sci. Bull. 1973, 18, 347–365. [Google Scholar
(https://scholar.google.com/scholar_lookup?
buscarmenú
title=Firm+Reservoir+Yield%E2%80%94How+Reliable+are+Historic+Hydrological+Rec
ords?
&author=Wallis,+J.R.&author=O%E2%80%99Connell,+P.E.&publication_year=1973&jo
urnal=Hydrol.+Sci.+Bull.&volume=18&pages=347%E2%80%93365&doi=10.1080/026
26667309494046)] [CrossRef (https://doi.org/10.1080/02626667309494046)]
68. Lettenmaier, D.P.; Burges, S.J. Operational assessment of hydrologic models of long-
term persistence. Water Resour. Res. 1977, 13, 113–124. [Google Scholar
(https://scholar.google.com/scholar_lookup?
title=Operational+assessment+of+hydrologic+models+of+long-
term+persistence&author=Lettenmaier,+D.P.&author=Burges,+S.J.&publication_year=
1977&journal=Water+Resour.+Res.&volume=13&pages=113%E2%80%93124&doi=10
.1029/WR013i001p00113)] [CrossRef (https://doi.org/10.1029/WR013i001p00113)]
69. McLeod, A.I.; Hipel, K.W. Preservation of the rescaled adjusted range: 1. A
reassessment of the Hurst Phenomenon. Water Resour. Res. 1978, 14, 491–508. [Goo‐
gle Scholar (https://scholar.google.com/scholar_lookup?
title=Preservation+of+the+rescaled+adjusted+range:+1.+A+reassessment+of+the+Hu
rst+Phenomenon&author=McLeod,+A.I.&author=Hipel,+K.W.&publication_year=1978
&journal=Water+Resour.+Res.&volume=14&pages=491%E2%80%93508&doi=10.102
9/WR014i003p00491)] [CrossRef (https://doi.org/10.1029/WR014i003p00491)]
70. Mandelbrot, B.B. Time varying channels, 1/f noises and the infrared catastrophe, or:
Why does the low frequency energy sometimes seem infinite? In Proceedings of the
the 1st IEEE Annual Communications Convention, Boulder, CO, USA, 7–9 June 1965.
[Google Scholar (https://scholar.google.com/scholar_lookup?
title=Time+varying+channels,+1/f+noises+and+the+infrared+catastrophe,+or:+Why+d
oes+the+low+frequency+energy+sometimes+seem+infinite?
&conference=Proceedings+of+the+the+1st+IEEE+Annual+Communications+Conventi
on&author=Mandelbrot,+B.B.&publication_year=1965)]
71. Watkins, modelos de renovación fraccional 1/ f de NW Mandelbrot de 1963-67: el
eslabón perdido no ergonómico entre los puntos de cambio y la dependencia de largo
alcance. arXiv , 2016; arXiv:1603.00738. [ Google Académico
(https://scholar.google.com/scholar_lookup?
title=Mandelbrot%E2%80%99s+1/f+fractional+renewal+models+of+1963-
67:+The+non-
ergodic+missing+link+between+change+points+and+long+range+dependence&autho
r=Watkins,+N.W.&publication_year=2016) ]
72. Scheidegger,
(/)
AE Modelos estocásticos en hidrología. Recurso Acuático. Res. 1970 , 6 ,
750–755. [ Google Scholar (https://scholar.google.com/scholar_lookup?

buscarmenú
title=Stochastic+Models+in+Hydrology&author=Scheidegger,+A.E.&publication_year=
1970&journal=Water+Resour.+Res.&volume=6&pages=750%E2%80%93755&doi=10.
1029/WR006i003p00750) ] [ CrossRef (https://doi.org/10.1029/WR006i003p00750) ]
73. Mandelbrot, BB Comentario sobre “Modelos estocásticos en hidrología”. Recurso
Acuático. Res. 1970 , 6 , 1791. [ Google Scholar
(https://scholar.google.com/scholar_lookup?
title=Comment+on+%E2%80%9CStochastic+Models+in+Hydrology%E2%80%9D&auth
or=Mandelbrot,+B.B.&publication_year=1970&journal=Water+Resour.+Res.&volume=
6&pages=1791&doi=10.1029/WR006i006p01791) ] [ CrossRef
(https://doi.org/10.1029/WR006i006p01791) ]

74. Wallis, JR; Matalas, NC Sensibilidad del diseño del yacimiento al mecanismo
generador de afluencias. Recurso Acuático. Res. 1972 , 8 , 634–641. [ Google Scholar
(https://scholar.google.com/scholar_lookup?
title=Sensitivity+of+reservoir+design+to+the+generating+mechanism+of+inflows&au
thor=Wallis,+J.R.&author=Matalas,+N.C.&publication_year=1972&journal=Water+Res
our.+Res.&volume=8&pages=634%E2%80%93641&doi=10.1029/WR008i003p00634)
] [ CrossRef (https://doi.org/10.1029/WR008i003p00634) ]
75. Klemeš, V.; Srikanthan, R.; McMahon, TA Modelos de flujo de memoria larga en el
análisis de yacimientos: ¿Cuál es su valor práctico? Recurso Acuático. Res. 1981 , 17 ,
737–751. [ Google Scholar (https://scholar.google.com/scholar_lookup?title=Long-
memory+flow+models+in+reservoir+analysis:+What+is+their+practical+value?
&author=Kleme%C5%A1,+V.&author=Srikanthan,+R.&author=McMahon,+T.A.&publica
tion_year=1981&journal=Water+Resour.+Res.&volume=17&pages=737%E2%80%937
51&doi=10.1029/WR017i003p00737) ] [ CrossRef
(https://doi.org/10.1029/WR017i003p00737) ]
76. Granger, CWJ Nuevas clases de modelos de series temporales. Estadística JR. Soc.
Ser. D 1978 , 27 , 237–253. [ Google Scholar
(https://scholar.google.com/scholar_lookup?
title=New+Classes+of+Time+Series+Models&author=Granger,+C.W.J.&publication_ye
ar=1978&journal=J.+R.+Stat.+Soc.+Ser.+D&volume=27&pages=237%E2%80%93253&
doi=10.2307/2988186) ] [ CrossRef (https://doi.org/10.2307/2988186) ]
77. Adenstedt,
(/)
RK sobre la estimación de muestras grandes de la media de una secuencia
aleatoria estacionaria. Ana. Estadística. 1974 , 2 , 1095-1107. [ Google Scholar

buscarmenú
(https://scholar.google.com/scholar_lookup?title=On+Large-
sample+Estimation+for+the+Mean+of+a+Stationary+Random+Sequence&author=Ade
nstedt,+R.K.&publication_year=1974&journal=Ann.+Stat.&volume=2&pages=1095%E
2%80%931107&doi=10.1214/aos/1176342867) ] [ CrossRef
(https://doi.org/10.1214/aos/1176342867) ]
78. Barnes, JA; Allan, DW Un modelo estadístico de ruido de parpadeo. Proc. IEEE 1966 ,
54 , 176–178. [ Google Scholar (https://scholar.google.com/scholar_lookup?
title=A+statistical+model+of+flicker+noise&author=Barnes,+J.A.&author=Allan,+D.W.
&publication_year=1966&journal=Proc.+IEEE&volume=54&pages=176%E2%80%9317
8&doi=10.1109/PROC.1966.4630) ] [ CrossRef
(https://doi.org/10.1109/PROC.1966.4630) ]
79. Hosking, JRM Diferenciación fraccionaria. Biometrika 1981 , 68 , 165-176. [ Google
Scholar (https://scholar.google.com/scholar_lookup?
title=Fractional+differencing&author=Hosking,+J.R.M.&publication_year=1981&journ
al=Biometrika&volume=68&pages=165%E2%80%93176&doi=10.1093/biomet/68.1.1
65) ] [ CrossRef (https://doi.org/10.1093/biomet/68.1.165) ]
80. Granger, CWJ; Joyeux, R. Introducción a los modelos de series temporales de memoria
larga y la diferenciación fraccionaria. J. Hora Ser. Anal. 1980 , 1 , 15-29. [ Google Scho‐
lar (https://scholar.google.com/scholar_lookup?title=An+Introduction+to+Long-
memory+Time+Series+Models+and+Fractional+Differencing&author=Granger,+C.W.J.
&author=Joyeux,+R.&publication_year=1980&journal=J.+Time+Ser.+Anal.&volume=1
&pages=15%E2%80%9329&doi=10.1111/j.1467-9892.1980.tb00297.x) ] [ CrossRef
(https://doi.org/10.1111/j.1467-9892.1980.tb00297.x) ]
81. Granger, CWJ Relaciones de memoria larga y agregación de modelos dinámicos. J.
Economía. 1980 , 14 , 227–238. [ Google Scholar
(https://scholar.google.com/scholar_lookup?
title=Long+Memory+Relationships+and+the+Aggregation+of+Dynamic+Models&auth
or=Granger,+C.W.J.&publication_year=1980&journal=J.+Econom.&volume=14&pages
=227%E2%80%93238&doi=10.1016/0304-4076(80)90092-5) ] [ CrossRef
(https://doi.org/10.1016/0304-4076(80)90092-5) ]
82. Taqqu,
(/)
MS Movimiento browniano fraccional y dependencia de largo alcance. En teoría
y aplicaciones de la dependencia a largo plazo ; Doukhan, P., Oppenheim, G., Taqqu, MS,

buscarmenú
Eds.; Birkhäuser: Boston, MA, EE.UU., 2003; págs. 5–38. [ Google Académico
(https://scholar.google.com/scholar_lookup?
title=Fractional+Brownian+Motion+and+Long-
Range+Dependence&author=Taqqu,+M.S.&publication_year=2003&pages=5%E2%80
%9338) ]
83. Mandelbrot, BB Dependencia global (a largo plazo) en economía y finanzas. En
Autoafinidad gaussiana y fractales: globalidad, la Tierra, ruido 1/f y R/S ; Springer: Nueva
York, Nueva York, Estados Unidos, 2002; págs. 601–610. [ Google Académico
(https://scholar.google.com/scholar_lookup?title=Global+(long-
term)+dependence+in+economics+and+finance&author=Mandelbrot,+B.B.&publicatio
n_year=2002&pages=601%E2%80%93610) ]
84. Imbers, J.; López, A.; Huntingford, C.; Allen, M. Sensibilidad de la detección del cambio
climático y atribución a la caracterización de la variabilidad climática interna. J. Clim.
2014 , 27 , 3477–3491. [ Google Scholar (https://scholar.google.com/scholar_lookup?
title=Sensitivity+of+Climate+Change+Detection+and+Attribution+to+the+Characteriz
ation+of+Internal+Climate+Variability&author=Imbers,+J.&author=Lopez,+A.&author
=Huntingford,+C.&author=Allen,+M.&publication_year=2014&journal=J.+Clim.&volu
me=27&pages=3477%E2%80%933491&doi=10.1175/JCLI-D-12-00622.1) ] [ Cross‐
Ref (https://doi.org/10.1175/JCLI-D-12-00622.1) ][ Versión verde (https://core.ac.uk/
download/pdf/19578547.pdf) ]
85. Kupferman, R. Cinética fraccionaria en modelos de baños térmicos de Kac-Zwanzig. J.
estadística. Física. 2004 , 114 , 291–326. [ Google Scholar
(https://scholar.google.com/scholar_lookup?title=Fractional+Kinetics+in+Kac-
Zwanzig+Heat+Bath+Models&author=Kupferman,+R.&publication_year=2004&journal
=J.+Stat.+Phys.&volume=114&pages=291%E2%80%93326&doi=10.1023/B:JOSS.00
00003113.22621.f0) ] [ CrossRef
(https://doi.org/10.1023/B:JOSS.0000003113.22621.f0) ]
86. Diebold, FX; Inoue, A. Memoria larga y cambio de régimen. J. Economía. 2001 , 105 ,
131-159. [ Google Scholar (https://scholar.google.com/scholar_lookup?
title=Long+memory+and+regime+switching&author=Diebold,+F.X.&author=Inoue,+A.
&publication_year=2001&journal=J.+Econom.&volume=105&pages=131%E2%80%93
159&doi=10.1016/S0304-4076(01)00073-2) ] [ CrossRef
(https://doi.org/10.1016/S0304-4076(01)00073-2) ]
87. Lo, AW
(/)
Memoria a largo plazo en los precios del mercado de valores. Econométrica
1991 , 59 , 1279-1313. [ Google Scholar (https://scholar.google.com/scholar_lookup?
title=Long-
buscarmenú
term+Memory+in+Stock+Market+Prices&author=Lo,+A.W.&publication_year=1991&jo
urnal=Econometrica&volume=59&pages=1279%E2%80%931313&doi=10.2307/2938
368) ] [ CrossRef (https://doi.org/10.2307/2938368) ]

88. Graves, T. Un enfoque sistemático de la inferencia bayesiana para procesos de


memoria larga. Doctor. Tesis, Universidad de Cambridge, Cambridge, Reino Unido,
2013. [ Google Scholar (https://scholar.google.com/scholar_lookup?
title=A+Systematic+Approach+to+Bayesian+Inference+for+Long+Memory+Processes
&author=Graves,+T.&publication_year=2013) ]

© 2017 por los autores. Licenciatario MDPI, Basilea, Suiza. Este artículo es un artículo de
acceso abierto distribuido bajo los términos y condiciones de la licencia Creative Commons
Attribution (CC BY) ( http://creativecommons.org/licenses/by/4.0/ (http://creativecommons.
org/licenses/by/4.0/) ).

Compartir y citar

 (mailto:?
&subject=From%20MDPI%3A%20%22A%20Brief%20History%20of%20Long%20Memory%3A
%20Hurst%2C%20Mandelbrot%20and%20the%20Road%20to%20ARFIMA%2C%201951%E2
%80%931980"&body=https://www.mdpi.com/218966%3A%0A%0AA%20Brief%20History%20
of%20Long%20Memory%3A%20Hurst%2C%20Mandelbrot%20and%20the%20Road%20to%2
0ARFIMA%2C%201951%E2%80%931980%0A%0AAbstract%3A%20Long%20memory%20pla
ys%20an%20important%20role%20in%20many%20fields%20by%20determining%20the%20b
ehaviour%20and%20predictability%20of%20systems%3B%20for%20instance%2C%20climat
e%2C%20hydrology%2C%20finance%2C%20networks%20and%20DNA%20sequencing.%20I
n%20particular%2C%20it%20is%20important%20to%20test%20if%20a%20process%20is%2
0exhibiting%20long%20memory%20since%20that%20impacts%20the%20accuracy%20and%
20confidence%20with%20which%20one%20may%20predict%20future%20events%20on%20
the%20basis%20of%20a%20small%20amount%20of%20historical%20data.%20A%20major%
20force%20in%20the%20development%20and%20study%20of%20long%20memory%20was
%20the%20late%20Benoit%20B.%20Mandelbrot.%20Here%2C%20we%20discuss%20the%2
0original%20motivation%20of%20the%20development%20of%20long%20memory%20and%
20Mandelbrot%E2%80%99s%20influence%20on%20this%20fascinating%20field.%20We%20
will%20also%20elucidate%20the%20sometimes%20contrasting%20approaches%20to%20lo
ng%20memory%20in%20different%20scientific%20communities.) 
(https://twitter.com/intent/tweet?
text=A+Brief+History+of+Long+Memory%3A+Hurst%2C+Mandelbrot+and+the+Road+to+AR
(/)
FIMA%2C+1951%E2%80%931980&hashtags=mdpientropy&url=https%3A%2F%2Fwww.mdpi
.com%2F218966&via=Entropy_MDPI)  buscarmen
( http://www.linkedin.com/shareArticle?
ú
mini=true&url=https%3A%2F%2Fwww.mdpi.com%2F218966&title=A%20Brief%20History%2
0of%20Long%20Memory%3A%20Hurst%2C%20Mandelbrot%20and%20the%20Road%20to%
20ARFIMA%2C%201951%E2%80%931980%26source%3Dhttps%3A%2F%2Fwww.mdpi.com
%26summary%3DLong%20memory%20plays%20an%20important%20role%20in%20many%2
0fields%20by%20determining%20the%20behaviour%20and%20predictability%20of%20syste
ms%3B%20for%20instance%2C%20climate%2C%20hydrology%2C%20finance%2C%20netwo
rks%20and%20DNA%20sequencing.%20In%20particular%2C%20it%20is%20important%20to
%20test%20if%20a%20process%20is%20exhibiting%20%5B...%5D) 
(https://www.facebook.com/sharer.php?u=https://www.mdpi.com/218966) 
(http://www.reddit.com/submit?url=https://www.mdpi.com/218966)
(http://www.mendeley.com/import/?url=https://www.mdpi.com/218966)

Estilo MDPI y ACS


Tumbas, T.; Gramacia, R.; Watkins, N.; Franzke, C. Una breve historia de la memoria larga:
Hurst, Mandelbrot y el camino hacia ARFIMA, 1951-1980. Entropía 2017 , 19 , 437.
https://doi.org/10.3390/e19090437

Estilo AMA
Graves T, Gramacy R, Watkins N, Franzke C. Una breve historia de la memoria larga: Hurst,
Mandelbrot y el camino hacia ARFIMA, 1951-1980. Entropía . 2017; 19(9):437.
https://doi.org/10.3390/e19090437

Estilo Chicago/Turabiano
Graves, Timothy, Robert Gramacy, Nicholas Watkins y Christian Franzke. 2017. "Una breve
historia de la memoria larga: Hurst, Mandelbrot y el camino hacia ARFIMA, 1951-1980"
Entropy 19, no. 9: 437. https://doi.org/10.3390/e19090437

error_outline Tenga en cuenta que desde el primer número de 2016, esta revista utiliza números de
artículos en lugar de números de páginas. Ver más detalles aquí
(https://www.mdpi.com/about/announcements/784) .

Métricas del artículo

Citas
anuncios referencia Web de la Scopus Google
cruzada Ciencia Académico
(/) 83 90 (https://
(https://ui.adsa GWVersion=2&partnerID=HzOxtitle=A+Brief+H

Estadísticas de acceso a artículos buscarmenú


Article access statistics
12.5k

10k

7.5k

5k

2.5k

0k
11 2 1 1 2 3 1 2 3
. S 1. S . Oc 1. O 1. O 1. O 0. N 0. N 0. N
ep ep t ct ct ct ov ov ov

Article Views

Para obtener más información sobre las estadísticas de la revista, haga clic aquí
(/journal/entropy/stats) .

error_outline Varias solicitudes de la misma dirección IP se cuentan como una vista.


Entropía (/journal/entropy) , EISSN 1099-4300, publicado por MDPI
RSS (/rss/journal/entropy) Alerta de contenido (/journal/entropy/toc-alert)

Más información

Cargos por procesamiento de artículos (/apc)


Pagar una factura (/about/payment)
Política de acceso abierto (/openaccess)
Contacto MDPI (/about/contact)
Empleos en MDPI (https://careers.mdpi.com)

Pautas

Para autores (/authors)


Para revisores (/reviewers)
(/)
Para editores (/editors)
Para bibliotecarios (/librarians)
Para editores (/publishing_services)
buscarmenú
Para Sociedades (/societies)
Para organizadores de conferencias (/conference_organizers)
Iniciativas MDPI

Sciforo (https://sciforum.net)
Libros MDPI (https://www.mdpi.com/books)
Preprints.org (https://www.preprints.org)
Scilit (https://www.scilit.net)
Perfiles de ciencia (https://sciprofiles.com?
utm_source=mpdi.com&utm_medium=bottom_menu&utm_campaign=initiative)
Enciclopedia (https://encyclopedia.pub)
ENJAMBRES (https://jams.pub)
Serie de actas (/about/proceedings)

Seguir MDPI

LinkedIn (https://www.linkedin.com/company/mdpi)
Facebook (https://www.facebook.com/MDPIOpenAccessPublishing)
Gorjeo (https://twitter.com/MDPIOpenAccess)

Suscríbase para recibir notificaciones


de publicación de números y
boletines informativos de las revistas
MDPI

Seleccione opciones

Introduce tu dirección de correo electróni

Suscribir

© 1996-2023 MDPI (Basilea, Suiza) a menos que se indique lo contrario

Descargo de responsabilidad Términos y condiciones (/about/terms-and-conditions)


política de privacidad (/about/privacy)

You might also like