You are on page 1of 21

EXPERIMENTS WITH A

SINGLE FACTOR:THE
A N A L Y S I S O F VA R I A N C E
THE ANALYSIS OF VARIANCE
Suppose we have a treatments or different levels of a single factor that we
wish to compare. The observed response from each of the a treatments is a
random variable.
Models for the Data

where yij is the ijth observation, μi is the mean of


the ith factor level or treatment, and ij is a
random error component. This equation is called
means model

μ is a parameter common to all treatments called


the overall mean,and τi is a parameter unique to
the ith treatment called the ith treatment effect.
This equation is usually called the effects model.
Both the means model and the effects model are linear statistical models; that is, the
response variable yij is a linear function of the model parameters

Both models are also called the one-way or single-factor analysis of variance
(ANOVA) model because only one factor is investigated.

Thus, the experimental design is a completely randomized design. Our objectives will
be to test appropriate hypotheses about the treatment means and to estimate them. For
hypothesis testing, the model errors are assumed to be normally and independently
distributed random variables with mean zero and variance 𝜎 2 .
Estadística  ANOVA  Un solo factor

To use a fixed significance level approach,


suppose that the experimenter has selected 0.05.
From Appendix Table IV we find that

F0.05,3,16 = 3.24.

Because F0 66.80 > 3.24,

we reject H0 and conclude that the treatment


means differ; that is, the RF power setting
significantly affects the mean etch rate
Gráfia  Gráfica de distribución de
probabilidad Gráfica de distribución
F; df1=3; df2=16
0,8

0,7

0,6

0,5

Densidad
0,4

0,3

0,2

0,1

0,01
0,0
0 5,292
X
Gráfica de intervalos de etch vs. power
95% IC para la media
750

700

650
etch

600

550

100 180 200 220


power
La desviación estándar agrupada se utilizó para calcular los intervalos.
A confidence interval

percent confidence interval on the ith treatment mean µi is

percent confidence interval on the difference in any two treatments means, say µi - µj,
would be
Medias

power N Media Desv.Est. IC de 95%


100 5 551,20 20,02 (533,88; 568,52)
180 5 587,40 16,74 (570,08; 604,72)
200 5 625,40 20,53 (608,08; 642,72) Apéndice II
220 5 707,00 15,25 (689,68; 724,32)
Informe de resumen de etch
Prueba de normalidad de Anderson-Darling
A-cuadrado 0,49
Valor p 0,192
Media 617,75
Desv.Est. 61,65
Varianza 3800,51
Asimetría 0,451923
Curtosis -0,978223
N 20
Mínimo 530,00
1er cuartil 571,25
Mediana 605,00
3er cuartil 676,50
Máximo 725,00
Intervalo de confianza de 95% para la media

540 570 600 630 660 690 720


588,90 646,60
Intervalo de confianza de 95% para la mediana
575,94 647,71
Intervalo de confianza de 95% para la desviación estándar
46,88 90,04

Intervalos de confianza de 95%

Media

Mediana

580 600 620 640


Estadísticas  ANOVA  Modelo
Lineal General

Almacenamiento  residuos
estandarizados

The general impression from


examining this display is that the
error distribution is approximately
normal. The tendency of the normal
probability plot to bend down
slightly on the left side and upward
slightly on the right side implies that
the tails of the error distribution are
somewhat thinner than would be
anticipated in a normal distribution;
that is, the largest residuals are not
quite as large (in absolute value) as
expected.
There is no reason to suspect any violation of the No unusual structure is apparent
independence or constant variance assumptions
Estadística  Estadística
básica  prueba de
normalidad
MODELO DE REGRESIÓN

Estadística  regresión  Gráfica de línea ajustada

Gráfica de línea ajustada Gráfica de línea ajustada


ETCH = 137,6 + 2,527 RFPOWER ETCH = 1148 - 8,255 RFPOWER
750 + 0,02837 RFPOWER^2
S 21,5413
R-cuad. 88,4% 750
S 18,4327
R-cuad.(ajustado) 87,8% R-cuad. 92,0%
R-cuad.(ajustado) 91,1%
700
700

650
ETCH

650

ETCH
600
600

550 550

160 170 180 190 200 210 220 160 170 180 190 200 210 220
RFPOWER RFPOWER
Comparing Treatment Means with a Control

Tukey test
Comparing Treatment Means with a Control

The Fisher Least Significant Difference (LSD) Method


Comparing Treatment Means with a Control

Dunnett´s procedure Appendix Table VIII.

You might also like