You are on page 1of 5

# MOMENT, SKEWNESS AND KURTOSIS

Moments
If X1, X2, .XN are the N Value assumsed by the variable X, we define the
quality
N

X r X 2r .......N Nr
Xr 1

r
j

i 1

(1)

## Called the moment. The first moment with r = 1 is arithmetic mean X

The rth moment about the mean X is defined as
N

mr =

(X

X )r

j 1

(X X )

(2)

( X X )r

## If r = 1, m1 = 0 (see Problem 3.16, Chap.3). If r = 2, m2 , the variance.

The rth moment about any origin A is defined as.
N

m
1
r

(X
j 1

A) r

( X A)

(3)

( X A) r

Where d=X A the deviations of X from A. If A = 0,(3) reduces to (1). For this reason (1)
is often called the rth moment about zero.
MOMENT FOR GROUPED DATA
If X1, X2, XR occured with frequencies f1, f2 fR respectively, the above
moment are given by
K

f X f 2 X ....... f k X
X 1
N
r
1

r
2

mr

f (X
j 1

N
K

m
1
r

X )r

f (X
j 1

A) r

f ( X A)

grouped data.

f
j 1

j 1

r
1

fX

(4)

(5)

f ( X X )r

Where N =

r
K

f X

( X A) r

(6)

## RELATIONS BETWEEN MOMENTS

The following relations exits between moment about the mean m r and moment
m2 = m2 m12
m3 = m3 3m1m2 + 2m13

(7)

## m4 = m4 4m1m3 + 6m12 m2 - 3m14

etc. see Problem 5.5) note that m1 = X A
COMPUTATION OF MOMENTS FOR GROUPED DATA

mr' = c r

f u
N

cr ur

(8)

## which can be used to find mr by applying equations (7)

COMPUTATION OF MOMENTS FOR GROUPED DATA
The coding method for computing the mean and standard deviation as given in
previous chapters can also be used to provide a short method for calculation of moments.
This method uses the fact that X j= A 4- cal (or briefly X = A + cu) so that from equation
(6) we have which can be used to find mr by applying equations (7).

## CHARLIER'S CHECK AND SHEPPARD'S CORRECTIONS

Charlier's check in computing moments by the coding method uses the identities :
f (u 1) fu N
f (u 1) 2 fu 2 2fu N
f (u 1)3 fu 3 3fu 2 3fu N

(9)

## f (u 1) 4 fu 4 4fu 4 6fu 2 4fu N

Sheppard's corrections for moments (extending the ideas on Page 72) are as follows :

Corrected m2 = m2

1 2
c
12

Corrected m4 = m4

1 2
7 4
c m2
c
12
240

## The moments ml and m3 need no correction.

MOMENTS IN DIMENSIONLESS FORM
To avoid particular units we can define the dimensionless moments about the mean
mr
ar = s
r

mr
m2

mr

(10)

m2r

## where s = is the standard deviation. Since m, = 0 and m2 = s2, we have a1 = 0, a2= 1.

SKEWNESS
Skewness is the degree of asymmetry, or departure from symmetry, of a
distribution. If the frequency curve (smoothed frequency polygon) of a distribution has a
longer "tail" to the right of the central maximum than to the left, the distribution is said to
be skewed to the right or to have positive skewness. If the reverse is true it is said to be
skewed to the left or to have negative skewness.
For skewed distributions the mean tends to lie on the same side of the mode as the
longer tail (see Figs. 3-1, 3-2, Chap. 3). Thus a measure of the asymmetry is supplied by
the difference (Mean Mode). This can be made dimensionless on division by a measure
of dispersion, such as the standard deviation, leading to the definition
Skewness

mean - mode
X mode

standard deviation
s

(11)

To avoid use of the mode, we can employ the empirical formula (10) on Page 48
and define :
Skewness

## 3(mean - median) 3( X median)

standard deviation
s

(12)

The above two measures are called, respectively, Pearson's first and second
coefficients of skewness.
Other measures of skewness defined in terms of quartiles and percentiles are as
follows:
Quartile coefficient of skewness =

Q3 Q1
Q3 Q1

(13)

P90 P10

## P90 2 P50 P10

(14)
P90 P10

An important measure of skewness uses the third moment about the mean
expressed in dimensionless form and is given by
m

3
3
3
Moment coefficient of skewness = a3 = s
3
( m2 )
m23
3

## Another measure of skewness is sometimes given by b1 = a3. For perfectly

symmetrical curves, such as the normal curve, a3 and 61 are zero.

KURTOSIS
Kurtosis is the degree of peakedness of a distribution, usually taken relative to a
normal distribution. A distribution having a relatively high peak such as the curve of Fig.
5-1(a) is called leptokurtic, while the curve of Fig. 5-1(b) which is flat-topped is called
platykurtic. The normal distribution, Fig. 5-1(c), which is not very peaked or very flattopped is called mesokurtic.

One measure of kurtosis uses the fourth moment about the mean expressed in
dimensionless form and is given by.
Moment coefficient of kurtosis = a4 =

m4 m4

s4 m22

(16)

which is often designated by 62. For the normal distribution, b2 =. a4 = 3. For this reason
the kurtosis is sometimes defined by (b2 3) which is positive for a leptokurtic
distribution, negative for a platykurtic distribution, and zero for the normal distribution.
Another measure of kurtosis which is also used is based on both quartiles and
percentiles and is given by
k

Q
(Q3 Q1)
P90 P10

(17)

Where Q =

1
(Q3 Q1) is the semi interquartile range. We shall refer to this to this
2

as the percentil confficient of kurtosis. For the normal distribution this has the value 0.026.
see Problem 5.14
POPULATION MOMENTS, SKEWNESS AND KURTOSIS
When it is required to distinguish moments and measures of skewness and kurtosis
for a sample from those corresponding to a population of which the sample is a part, it is
often the custom to use Latin symbols for the former and Greek symbols for the latter.
Thus if sample moments are denoted by m r and mr' the corresponding Greek symbols
would berand lir' (II is the Greek letter mu). Subscripts are always denoted by Latin
symbols.
Similarly if the sample measures of skewness and kurtosis are denoted by a, and a4
respectively, the population skewness and kurtosis would be a 3 and a, (a is the Greek letter
alpha).
We have already mentioned that the standard deviation for sample and population
are denoted respectively by s and