3 Dispersion Skewness Kurtosis PDF

Descriptive Statistics II
Dr. Umesh R A
Agenda
• Measures of Dispersion
• Skewness
• Kurtosis
Is Central Tendency of data is enough to
describe the frequency distribution?
Yes or No?
What is Measures of Dispersion (Variation)?
Why Measures of Dispersion?

Dispersion
Dispersion:
The property of deviation of values from the
average is called variation or dispersion.
Measures of Variations:
The degree of variation is indicated by Measures
of Variations.
Dispersion
Measures of Variation:
• Range
• Interquartile Range
• Mean Deviation (M.D.)
• Standard Deviation (S.D.)
The above measures are absolute measures of

variation.
Relative measures of variation
However, for comparison of variation in two or
more frequency distribution, a relative measure
of variation should be calculated.
Relative measures of variation:
1. Coefficient of Range
2. Coefficient of Quartile Deviation
3. Coefficient of Mean Deviation
4. Coefficient of Variation (C.V.)
Absolute vs Relative Measures
• The absolute measures are dependent of the
units of Measurement. So, we can not
compare two different frequency distributions
directly.
• The relative measures are independent of the

units of Measurement. And so, they facilitate
comparison.
Requisites of Good Measure of
Dispersion
• It should be based on all the observations.
• It should not be affected by extreme value.
• It should be rigidly defined.
• It should be easy to calculate and understand.
• It should be capable of further algebraic
treatment.
• It should be Possible to find for open end class
intervals.
Range
Range is the difference between the highest
and the lowest value in the data.
𝑅 =𝐻−𝐿
where, H-Highest Value, L-Lowest Value
Coefficient of range:
𝐻−𝐿
𝐶𝑜𝑒𝑓. 𝑜𝑓 𝑅𝑎𝑛𝑔𝑒 =
𝐻+𝐿
Example:
Range
• A major drawback of Range is that, since it is
based on extreme values, it is highly affected
by abnormal values.
• This drawback is rectified in quartile deviation.

Quartile Deviation (Q.D.)
Quartile Deviation (or Semi-Interquartile
Range)
• It is based on only lower and upper Quartiles.

• It is not based on all the observations.
• So, it is not affected by abnormal extreme
values.
Q.D.: Formulae
Quartile Deviation:
𝑄3 − 𝑄1
𝑄. 𝐷. =
2
• Coefficient of Quartile deviation:
𝑄3 − 𝑄1
𝐶𝑜𝑒𝑓. 𝑄. 𝐷. =
𝑄3 + 𝑄1
• Interquartile Range (IQR):
𝐼𝑄𝑅 = 𝑄3 − 𝑄1
Mean Deviation
• The mean deviation of a set of values from a
central value is the mean of absolute
deviations of the values from the central value
(Mean or Median or Mode).
Mean deviation about AM
For Raw data:
M.D.( X ) =
 x−x
n
For tabulated data:
M.D.( X ) =
 f x−x
N
Relative Measure:
M.D.( X )
Coeff of M.D.( X ) =
X
Mean deviation about Median
For Raw data:
M.D.(M) =
 x−M
n
For tabulated data:
M.D.(M) =
 f x−M
N
Relative Measure:
M.D.(M)
Coeff of M.D.(M) =
M
Tip! (about Mean deviation)
Mean deviation can be calculated about any
average - mean, mode, median, G.M., H.M.,
etc.
But, mean deviation is the least when it is

measured from the median (Minimal property
of Median).
Therefore, M.D.(M) should be preferred.

Standard Deviation (S.D.)
• The standard deviation of a set of values is the
positive square-root of mean of the squared
deviations of the values from their Arithmetic
Mean.
• It is denoted by σ (sigma).
S.D.: Formulae
Raw data Tabulated data
For
=  ( x − x ) 2
=  f ( x − x )2
Population n N
For =  ( x − x ) 2
=  f ( x − x )2
Sample n −1 N −1
Properties
1. S.D. is independent of the origin of
measurement but not independent of scale.
2. Always, σ ≥ 0.
3. S.D. is the least of all root-mean-square
deviations.
Note: SD= σ ,Variance= σ2

Coefficient of Variation (C.V.)
• Coefficient of Variation is relative measure of
Standard deviation.
Standard Deviation
C.V. = 100
Arithmetic Mean

C.V. = 100
x
• It is independent of units of measurement. So,
useful while comparing variation present in
different frequency distributions.
Is Central Tendency and variation enough to
describe frequency distributions?
Yes or No?
• Clearly NO.
• There few measures need to be computed to
describe data, such as;
– Skewness
– Kurtosis
• To compute skewness and kurtosis we need
to know the “Moments”.
Lets see!
Moments
• The characteristics of a frequency
distributions are described by its Moments.
• Definition: The rth moment of the set of values
about any constant is the mean of rth powers
of the deviation of the values from the
constant.
• Types of Moments:
– Central Moments
– Raw moments
Moments: Formulae
Here, ‘a’ is any constant except AM.
Central Moments Raw Moments

Raw data σ(𝑥 − 𝑥)ҧ 𝑟 σ(𝑥 − 𝑎) 𝑟
𝜇𝑟 = 𝜇𝑟′ =
𝑛 𝑛
Tabulated σ 𝑓(𝑥 − 𝑥)ҧ 𝑟 ′ σ 𝑓(𝑥 − 𝑎)𝑟
data 𝜇𝑟 = 𝜇𝑟 =
𝑁 𝑁
Moments
• Moments helps in measuring the
scatteredness, asymmetry and peakedness of
a curve for a particular distribution.
Applications of Moments
• The first moment about zero is the mean.
• The second moment about the mean is the
sample variance.
• The third moment about the mean in
calculating skewness.
• The fourth moment about the mean in the
calculation of kurtosis.
Skewness
Skewness means non-symmetry or asymmetry
or lack of symmetry.
Skewness
If mean = median = mode,
the shape of the distribution is symmetric.
If mode < median < mean,

the shape of the distribution trails to the right,
is positively skewed.
If mean < median < mode,

the shape of the distribution trails to the left, is
negatively skewed.
Measures of Skewness
1. Based on Moments
2. Karl Pearson’s Coefficient of Skewness
3. Bowley’s Coefficient of Skewness
1. Based on Moments
Formula:
𝜇32
𝛽1 = OR 𝛾1 = + 𝛽1
𝜇23
• Measure based on moments (i.e. 𝛽1 ) is more

exact in indicating degree of skewness of
skewness than others.
2. Karl Pearson’s Coefficient of
Skewness
Formula:
𝑚𝑒𝑎𝑛 − 𝑚𝑜𝑑𝑒 𝑥ҧ − 𝑍
𝑆= =
𝑆. 𝐷. 𝜎
Note:
If mode is ill-defined, we can go for following formula;
3(𝑚𝑒𝑎𝑛 − 𝑚𝑒𝑑𝑖𝑎𝑛) 3(𝑥ҧ − 𝑀)
𝑆= =
𝑆. 𝐷. 𝜎
since we know, 𝑍 = 3𝑀 − 2𝑥ҧ .
3. Bowley’s Coefficient of
Skewness
Bowley’s Coefficient of Skewness is based on
quartiles. Also called as Galton skewness.
Formula:
𝑄3 − 𝑄2 − (𝑄2 − 𝑄1 )
𝑆=
𝑄3 − 𝑄2 + (𝑄2 − 𝑄1 )
Kurtosis
Kurtosis is the degree of peakedness (non-
flatness).
Kurtosis: Based on Moments
Formula:
𝜇4
𝛽2 = OR 𝛾1 = 𝛽2 − 3
𝜇22
If, Decision If,

𝛽2 = 3 Platykurtic 𝛾1 = 0
𝛽2 < 3 Mesokurtic 𝛾1 < 0
𝛽2 > 3 Leptokurtic 𝛾1 > 0
Skewness
• If skewness = 0, the data are perfectly
symmetrical. But a skewness of exactly zero is
quite unlikely for real-world data, so how can
you interpret the skewness number?
Skewness
Bulmer, “Principles of Statistics” (1979) suggests
the following rule of thumb:
• If skewness is less than −1 or greater than +1,
the distribution is highly skewed.
• If skewness is between −1 and −½ or between
+½ and +1, the distribution is moderately
skewed.
• If skewness is between −½ and +½, the
distribution is approximately symmetric.
Example:
Question: In frequency distribution, first four
central moments are 0, 4, -2 and 2.4. Comment
on Skewness and Kurtosis of the distribution.
Solution:
Given, 𝜇1 = 0, 𝜇2 =4, 𝜇3 = −2 and 𝜇4 = 2.4
Therefore,
𝜇32 𝜇4
𝛽1 = = 0.0625 and 𝛽2 = = 0.15
𝜇23 𝜇22
Since, 𝝁𝟑 is negative, the distribution is negatively
skewed. Also, 𝜷𝟏 = 𝟎. 𝟎𝟔𝟐𝟓 is very small, the
distribution is slightly skewed.
Since, 𝜷𝟏 = 𝟎. 𝟏𝟓 < 𝟑, the distribution is platykurtic.

Next Lecture
•Topic: Probability
–Introduction to Probability
–Conditional probability
–Bayes Theorem
•Background material to study

–Business Statistics, Chpt2, pp. 50
•We will have MCQ on this lecture: 10 Questions
•And will have recap of the lecture one.
•Discuss on the assignment of this lecture.
Thank You

3 Dispersion Skewness Kurtosis PDF

Uploaded by

Copyright:

Available Formats

3 Dispersion Skewness Kurtosis PDF

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

3 Dispersion Skewness Kurtosis PDF

Uploaded by

Copyright:

Available Formats

Descriptive Statistics II

Why Measures of Dispersion?

The above measures are absolute measures of

• The relative measures are independent of the

• This drawback is rectified in quartile deviation.

• It is based on only lower and upper Quartiles.

But, mean deviation is the least when it is

Therefore, M.D.(M) should be preferred.

Note: SD= σ ,Variance= σ2

Central Moments Raw Moments

If mode < median < mean,

If mean < median < mode,

• Measure based on moments (i.e. 𝛽1 ) is more

If, Decision If,

Since, 𝜷𝟏 = 𝟎. 𝟏𝟓 < 𝟑, the distribution is platykurtic.

•Background material to study

You might also like