Normality Test
Normality Test
If there are no outliers, you might try a transformation (such as, the log or
square root) to make the data normal. If a transformation is not a viable
Alternative, nonparametric methods that do not require normality may be
used.
1
SEVEN DIFFERENT NORMALITY TESTS
SHAPIRO-WILK W TEST
This test for normality has been found to be the most powerful test in most
situations. It is the ratio of two estimates of the variance of a normal
distribution based on a random sample of n observations. The numerator is
proportional to the square of the best linear estimator of the standard
deviation. The denominator is the sum of squares of the observations about
the sample mean. The test statistic W may be written as the square of the
Pearson correlation coefficient between the ordered observations and a set of
weights which are used to calculate the numerator. Since these weights are
asymptotically proportional to the corresponding expected normal order
statistics, W is roughly a measure of the straightness of the normal quantile-
quantile plot. Hence, the closer W is to one, the more normal the sample is.
The probability values for W are valid for sample sizes greater than 3. The test
was developed by Shapiro and Wilk (1965) for sample sizes up to 20. NCSS
uses the approximations suggested by Royston (1992) and Royston (1995)
which allow unlimited sample sizes. Note that Royston only checked the
results for sample sizes up to 5000, but indicated that he saw no reason larger
sample sizes should not work. W may not be as powerful as other tests when
ties occur in your data. The test is not calculated when a frequency variable is
specified.
ANDERSON-DARLING TEST
2
MARTINEZ-IGLEWICZ TEST
This test for normality, developed by Martinez and Iglewicz (1981), is based
on the median and a robust estimator of dispersion. The authors have shown
that this test is very powerful for heavy-tailed symmetric distributions as well
as a variety of other situations. A value of the test statistic that is close to one
indicates that the distribution is normal. This test is recommended for
exploratory data analysis by Hoaglin (1983). The formula for this test is:
KOLMOGOROV-SMIRNOV TEST
This test for normality is based on the maximum difference between the
observed distribution and expected cumulative-normal distribution. Since it
uses the sample mean and standard deviation to calculate the expected
normal distribution, the Lilliefors’ adjustment is used. The smaller the
maximum difference the more likely that the distribution is normal.
This test has been shown to be less powerful than the other tests in most
situations. It is included because of its historical popularity.
3
D’AGOSTINO KURTOSIS TEST
D’Agostino (1990) describes a normality test based on the kurtosis coefficient,
√ b2. Recall that for the normal distribution, the theoretical value of √ b2 is 3.
Hence, a test can be developed to determine if the value of √ b2 is significantly
different from 3. If it is, the data are obviously non-normal. The statistic, Z k, is,
under the null hypothesis of normality, approximately normally distributed
for sample sizes n>20. The calculation of this test proceeds as follows:
4
D’AGOSTINO OMNIBUS
D’Agostino (1990) describes a normality test that combines the tests for
skewness and kurtosis. The statistic, K2, is approximately distributed as a chi-
square with two degrees of freedom. After calculated zs and zk, calculate K2 as
follows:
5
PROPERTIES OF NORMAL PROBABILITY CURVE
The Normal Probability Curve (N.P.C.) is symmetrical about the ordinate of the central
point of the curve. It implies that the size, shape and slope of the curve on one side of the
That is, the normal curve has a bilateral symmetry. If the figure is to be folded along its
vertical axis, the two halves would coincide. In other words the left and right values to the
middle central point are mirror images.
Since there is only one point in the curve which has maximum frequency, the normal
The mean, median and mode of the normal distribution are the same and they lie at the
centre. They are represented by 0 (zero) along the base line. [Mean = Median = Mode]
6
4. The maximum ordinate occurs at the centre:
The maximum height of the ordinate always occurs at the central point of the curve that is,
at the mid-point. The ordinate at the mean is the highest ordinate and it is denoted by Y 0.
(Y0 is the height of the curve at the mean or mid-point of the base line).
The Normal Probability Curve approaches the horizontal axis asymptotically i.e., the curve
continues to decrease in height on both ends away from the middle point (the maximum
It extends infinitely in both directions i.e. from minus infinity (-∞) to plus infinity (+∞) as
shown in Figure below. As the distance from the mean increases the curve approaches to
In the normal probability curve the height declines symmetrically in either direction from
the maximum point. Hence the ordinates for values of X = µ ± K, where K is a real number,
are equal.
For example:
The heights of the curve or the ordinate at X = µ + σ and X = µ – σ are exactly the same
7
7. The points of Influx occur at point ± 1 Standard Deviation (± 1 a):
The normal curve changes its direction from convex to concave at a point recognized as
point of influx. If we draw the perpendiculars from these two points of influx of the curve
on horizontal axis, these two will touch the axis at a distance one Standard Deviation unit
8. The total percentage of area of the normal curve within two points of influxation is
fixed:
Approximately 68.26% area of the curve falls within the limits of ±1 standard deviation
8
9. Normal curve is a smooth curve:
The normal curve is a smooth curve, not a histogram. It is moderately peaked. The kurtosis
The 50% area of the curve lies to the left side of the maximum central ordinate and 50%
The curve is used as a measurement scale. The measurement unit of this scale is ± σ (the
There is a greater percentage of cases at the middle of the distribution. In between -1σ and
+ 1σ, 68.26% (34.13 + 34.13), nearly 2/3 of eases lie. To the right side of +1σ, 15.87%
(13.59 + 2.14 + .14), and to the left of-1σ, 15.87% (13.59 + 2.14 + .14) of cases lie. Beyond
+2σ. 2.28% of eases lie and beyond -2σ also 2.28% of cases lie.
Thus, majority of eases lie at the middle of the distribution and gradually number of cases
Percentage of cases between Mean and different a distances can be read from the
figure below:
9
13. The scale of X-axis in normal curve is generalised by Z deviates
x = scores (expressed as deviations from the mean) laid off along the base line or X-axis.
y = the height of the curve above the X axis, i.e., the frequency of a given x-value.
N = number of eases
15. The normal curve is based on elementary principles of probability and the other name
10
NORMAL PROBABILITY CURVE : COMPUTATION,
CHARACTERISTICS AND APPLICATIONS
If a coin is tossed unbiased it will fall either head (H) or tail (T). This the probability of
Likewise of we shall toss two coins, coin x and coin y there are four possible ways of falling.
Thus the four possible ways are-both x and y may fall H, x may fall T and y H, x may fall H
Expressed in ratios
ADVERTISEMENTS:
The expected appearance of heads and tails of two coins can be expressed as:
If we shall increase the number of coins to three i.e. x, y and Z, there can be eight possible
arrangements.
11
The expected appearance of heads and tails of coins can be expressed as:
In this way we can determine the probability of different combinations of heads and tails of
any number of coins. We can obtain probability of any number of coins by binomial
theorem is an algebraic formula which expands the power of a binomial expression in the
form of a series.
n! means 1 x 2 x 3 x …. x n
12
If the above data are plotted on a graph as histogram and frequency polygon it will
13
Thus the figure we obtained from toss of 10 coins (H + T)10 is a symmetrical many sided
polygon.
And if we shall go on increasing the number of coins, with each increase the polygon
would exhibit a perfectly smooth surface line the figure-11.2 given below:
This bell shaped curve is called as the ‘Normal Probability Curve’. Thus the “graph of the
14
In statistics it is important because:
(а) It is the distribution of many naturally occurring variables, such as intelligence of 8th
(b) The distribution of the means of samples drawn from most parent populations is
Therefore normal curve has great significance in social sciences and behavioural sciences.
So that Normal Probability Curve or most popularly known as NPC is used as a reference
curve. In order to understand the utility of the NPC we must have to understand the
15
CHARACTERISTICS OF NORMAL PROBABILITY CURVE
The curve is symmetrical to its ordinate of the central point of the curve. It means the size,
shape and slope of the curve on one side of the curve is identical to the other side of the
curve. If the curve is bisected then its right hand side completely matches to the left hand
side.
The Normal Probability Curve approaches the horizontal axis and extends from-∞ to + ∞.
Means the extreme ends of the curve tends to touch the base line but never touch it.
The mean, Median and mode fall at the middle point and they are numerically equal.
The points of influx in a NPC occur at ± 1σ to unit above and below the mean. Thus at this
point the curve changes from convex to concave in relation to the horizontal axis.
16
5. The total area of NPC is divided in to ± standard deviations:
The total of NPC is divided into six standard deviation units. From the center it is divided in
to three +ve’ standard deviation units and three —ve’ standard deviation units.
Thus ± 3σ of NPC include different number of cases separately. Between ± 1σ lie the middle
2/3rd cases or 68.26%, between ± 2σ lie 95.44% cases and between ± 3σ lie 99.73% cases
The Y ordinate of the NPC represents the height of the curve. At the center the maximum
ordinate occurs. The height of the curve at the mean or mid point is denoted as Y 0.
In order to determine the height of the curve at any point we use the following
formula:
17
7. It is unimodal:
The curve is having only one peak point. Because the maximum frequency occurs only at
one point.
The height of the curve decline to both the direction symmetrically from the central point.
Means the M + σ and M — σ are equal if the distance from the mean is equal.
As the mean of the NPC represent the population mean so it is represented by the µ (Meu).
10. In Normal Probability Curve the Standard deviation is the 50% larger than the Q:
1 PE = .6745σ
1σ = 1.4826PE.
11. Q can be used as a unit of measurement in determining the area within a given
part:
18
12. The Average Deviation about the mean of NPC is .798σ:
NPC.
In a Normal Probability curve the modal ordinate varies increasingly to the standard
deviation. The standard deviation of the Normal Probability Curve increases, the modal
19
APPLICATIONS OF NORMAL PROBABILITY CURVE
Some of the most important applications of normal probability curve are as follows:
The principles of Normal Probability Curve are applied in the behavioural sciences in many
different areas.
given limits:
Example:
Solution:
Here first of all we have to convert both the scores 16 and 32 into a standard score.
Entering in to the Table-A, the table area under NPC, it is found that 34.13 cases fall
between mean and – 1σ and 34.13 cases fall between mean and + 1σ. So ± σ covers 68.26%
20
Example:
percentage of cases will lie above and below the score 36.
Solution:
First of all we have to convert the raw score 36 into standard score.
Entering into the Table-A, the table area under the NPC it is found that 19.15% cases fall
between Mean and -.5σ. Therefore the total percentage of cases above the score 36 is 50 +
19.15 = 69.15% and below the score 36 is 50-19.15 = 30.85%. So in the distribution
69.15% cases are above the score 36 and 30.85% scores are below the score 36.
2. NPC is used to determine the value of a score whose percentile rank is given:
By using NPC table we can determine the raw score of the individual if the percentile rank
is given.
Example:
In a distribution of scores of a doss Pinky’s percentile rank in statistics is 65. The mean of
the distribution is 55 with a standard deviation of 10. Find but the raw score of Pinky in
Statistics.
Solution:
As Pinky’s percentile rank is 65 so in a normal distribution her position is 35% above the
mean. By entering in to the table ‘A’ we found that 35% from the mean is + 1.04 σ.
21
3. NPC is used to find the limits in a normal distribution which include a given
percentage of cases:
When a distribution is normally distributed and what we know about the distribution is
Mean and the Standard deviation at that time by using the table area under NPC we can
Example:
Solution:
In a normal distribution the middle 75% cases include 37.5% cases above the mean and
37.5% cases below the mean. From the Table-A we can say that 37.5% cases covers 1.15 σ
units. Therefore the middle 75% cases lie between mean and ± 1.15 σ units.
So in this distribution middle 75% cases will include the limits 14.25 to 25.75.
If scores of two groups on a particular variable are normally distributed. What we know
about the group is the mean and standard deviation of both the groups. And we want to
22
know how much the first group over-laps the second group or vice-versa at that time we
5. NPC helps us in dividing a group into sub-groups according to certain ability and
When we want to divide a large group in to certain sub-groups according to some specified
ability at that time we use the standard deviation units of a NPC as units of scale.
Example:
An achievement test was administered to the 600 8th grade students. The teacher wants to
the test. Assuming the normality of the distribution of scores calculate the number of
Solution:
23
6. NPC helps to determine the relative difficulty of test items or problems:
When it is known that what percentage of students successfully solved a problem we can
determine the difficulty level of the item or problem by using table area under NPC.
In order to normalize a frequency distribution we use Normal Probability Curve. For the
In an experiment we test the relationship among variables whether these are due to chance
fluctuations or errors of sampling procedure or it is real relationship. This is done with the
24
9. NPC is used to generalize about population from the sample:
We compute standard error of mean, standard error of standard deviation and other
statistics to generalize about the population from which the sample are drawn. For this
25
SELF OPINION
We usually apply normality test to the results of processes that, under the null, generate
random variables that are only asymptotically or nearly normal (with the asymptotically
In the era of cheap memory, big data and fast processors, normality tests should always
reject the null of normal distribution for large (though not insanely large) samples. And so,
perversely, normality tests should only be used for small samples, when they presumably
In this way I have completed this practical of normality test and thankful to Dr. Gaikwad Sir
26
27