Normal Distribution

A normal distribution is one in which the majority of cases are located in the middle

of the scale and only a small number of cases are located at both extremes of the

scale. A normal curve is one, which graphically represents normal distribution. Most

of the traits or variables in psychology, sociology and education are normally

distributed. In reality, however, no distribution ever takes the absolute form of normal

distribution. Many frequency distributions are close to this absolute form and we

assume that they are normally distributed.

Characteristics of a normal probability curve

A normal probability curve is a bell shaped curve. The major characteristics of a normal

probability curve are

 In a normal probability curve the X axis represents the trait values as “z”

scores and Y axis represents the frequency in terms of percentage,

proportion etc.

 A normal curve is always symmetrical, that is, the right of the curve is

equivalent to the left of the curve.

 In a normal curve the mean, the median and the mode are numerically

identical and fall at the center of the distribution.

 A normal curve never touches the base line, no matter how far the curve is

stretched.

 In a normal curve the highest ordinate is at the center. All ordinates on both

sides of the distribution are smaller than the highest ordinate.

 A normal curve is continuous.

Area under the normal curve

104
 The scores of the trait under measurement are converted to “z” scores. It is

obtained by the formula z= x-m/ σ

Where x=the raw score, m=mean and σ = standard deviation.

The range of the normal probability curve (z scores) is from –3σ to +3σ with

0 representing the midpoint (mean) of the normal probability curve.

Figure-8.1: Different percentages of area under the normal curve

In a normal probability curve the range between –1σ to +1σ where one

standard deviation unit taken on each side of the mean includes a total area of

34.13% + 34.13% = 68.26% of the curve. The range between –2σ to +2 σ includes

the area 47.72% + 47.72%= 95.44%. The range between +3σ to –3σ includes the

area 49.865% + 49.865% = 99.73%. this shows that where as the area between 0 to

1σ is 34.13%, the area between 1σ and 2σ is 13.59% and the area between 2σ and

3σ is 2.15% indicating that the maximum concentration of scores are there near the

mean and as we go farther in positive or negative directions the concentration of

scores reduce. While referring to the normal probability tables it should be noted

that all the areas given are for the z scores from mean that is 0.

8.4. Deviations of the normal probability curve

Not all distributions are normally or nearly normally distributed. The deviations

105
to the normal distribution may occur in two ways. (1) skewness (2) Kurtosis.

8.4.1. Skewness : The distribution is said to be “skewed” when the mean and the

median fall at different point in the distribution, and the balance (or center of gravity)

is shifted to one side or the other (to left or right). In a normal distribution, the mean

equals the median exactly and the skewness is, of course, zero.

Distributions are said to be skewed negatively or to the left when scores are

massed at the high end of the scale (the right end) and are spread out more

gradually to ward the low end (the left) as shown in figure-8.2

Median
Mean an
Figure 8.2: Negative skewness: to the left

Distributions are skewed positively or to the right when scores are massed at

the low (or left) end of the scale, and are spread out gradually toward the high or

right end as shown in figure.8.3.

Median Mean
Figure.8.3. Postitive skewness: to the right

The greater the gap between mean and median, the greater the skewness.

106
When skewness is negative, the mean lies to the left of the median i.e. mean value

is lower than the median value; and when skewness is positive, the mean lies to the

right of the median i.e. the mean value higher than the median value.

The skewness of the distribution is obtained by the formula

Skewness (Sk) = 3 (mean – median) / σ

8.4.2. Kurtosis

The term “kurtosis” refers to the “peakness” or flatness of a frequency

distribution as compared with the normal. A frequency distribution more peaked than

the normal is said to be leptokurtic; one flatter than the normal, platykurtic. Figure-

8.4. shows a leptokurtic distribution and a platykurtic distribution plotted on the same

diagram around the same mean. A normal curve (called mesokurtic) also has been

drawn in on the diagram to bring out the contrast in the figure, and to make

comparison easier.

Figure.8.4. Leptokurtic (A), Normal or mesokurtic (B) and Platykurtic (C) curves

8.5. Applications of normal probability curve

A normal curve helps in testing the significance of the obtained measures

against a chance hypothesis and thus enables the researcher to make a

generalization about the population from which the sample was drawn.

8.6. Standard Error

107
 The primary objective of statistical inference is to enable us to generalize from

a sample to some larger population of which the sample is a part. A good sample is

assumed to be representative of the population. However most of the times the

measures obtained from the sample may not be identical to the population

measures. The sample measures may be slightly larger or smaller than the

population values. The degree to which the sample measures represent the

population measures is an index of the significance or trustworthiness of the

computed sample measures. Any deviation of such sample measure from the

population may be regarded as an error of estimation the standard error tells us how

large these errors of estimations are in any particular sampling situations. Therefore

the standard error of any measure is a standard deviation of the distribution of

sample measures. To distinguish such a standard deviation from the more familiar

one that applies to dispersions of individual observations it is called standard error

and is referred by the abbreviation SE.

For example let us assume mean is the statistic to be computed from a

sample drawn from a population. Instead of one sample if we draw number of

samples from the population and compute means for all these samples the standard

deviation of these means is the standard error of mean for the samples drawn from

the population.

Since it is not practically feasible to draw number of samples from the

population a standard error is computed from one sample drawn from the

population.

Every statistic computed should be accompanied by the standard value of

that statistic. In this chapter we shall see the computation of standard error for

measures of central tendency and variability.

108
8.6.1. Standard error of mean
The formula for the standard error of the mean is
σ
SEM or σM =
N
σ = the standard deviation of the population
N = the number of cases in the sample.

8.6.2. Standard error of median

The formula for the standard error of the median is

1.858Q
SE Mdn =
N

8.6.3. Standard error of standard deviation

the fomulat for the standard error of the standard deviation is

.71σ
SEσ =
N

8.6.4. Standard error of quartile deviation

The formula for the standard error of the quartile deviation is

1.71Q
SEQ =
N

8.6.5.Example of standard error of mean: The mean on a test of abstract

reasoning for 225 boys in the tenth class of a city was 27.26 with a SD of 11.20.

How dependable is this mean? Specifically, how good an estimate is it of the mean

which could be expected if all of the tenth class boys in city were tested.

From the formula we find that the SEM is

109
 11.20
SEM = = .75
225

The SEM obtained above can be thought of as a standard deviation of a distribution

of number of sample means around the fixed population mean. The normal curve in

the figure shows this sampling distribution. At its center (unknown) is the population

mean and its SD is 0.75. The sample mean may fall equally often on the positive

and negative sides of the population mean. A 0.05 level of significance means to

obtain a range where the sample mean is likely to be 47.5% on the positive side of

the population mean and 47.5% on the negative side of the population mean and the

chance factor is likely to occur 2.5% on positive side and 2.5% on the negative side.

From the normal probability tables 47.5% corresponds to +1.96 σ and -1.96 σ. It

means that 95 times out of 100 times the sample mean will not miss the population

mean by + 1.96 X 0.75 i.e. 1.47. Therefore the 95% confidence interval for the

population mean is 27.26 + 1.47, which is 25.79 and 28.83. So it means that if all

the tenth class students of the city were tested we can be confident 95 out of 100

times that it will be between 25.79 and 28.83. Similarly 99% confidence interval can

be obtained by looking into the normal probability tables for + 49.5 which is + 2.58.

Hence every statistic should be accompanied by its standard error to assess

the dependability of the obtained statistical value.

8.7. Significance

Whether a difference is to be taken as statistically significant or not depends

upon the probability that the given difference could have arisen “by chance.”

110
 Experimenters and research workers have for convenience chosen several

arbitrary standards called levels of significance of which the .05 and .01 levels are

most often used. A 0.05 level of significance means that the chance factor exists

2.5% on the positive side and 2.5% on the negative side of the normal probability

curve and 95% of the time the statistical value obtained is true. A 0.01 level of

significance means that the chance factor exists 0.5% on the positive side and 0.5%

on the negative side of the normal probability curve and 99% of the time the

statistical value obtained is true. The confidence with which an experimenter rejects

or retains a null hypothesis depends upon the level of significance adopted.

8.8. Null hypothesis

Experimenters have found the null hypothesis is a useful tool in testing significance.

In its simplest form this hypothesis asserts that there is no true difference between

two population means and that any difference found between sample means is

therefore accidental and unimportant. The null hypothesis is similar to legal principle

that a man is innocent until he is proved. It is a challenge to the experimenter to give

facts to reject null hypothesis.

8.9. One tail test and two tail test

A one tailed test is a directional test, which indicates the direction of the

difference between the samples under study. Here the change can occur only in one

direction. If we are interested in studying the language development from infancy

onwards the change is always in the positive direction which is the language

develops with the progress of each year. Here we are concerned whether the

change is significant or not. A two-tail test is one where the rejection of the null

hypothesis can be in either direction. While comparing two teaching methods A and

B if it is found that there is no difference the null hypothesis is accepted. But if the

111
null hypothesis is to be rejected it may be found that method A is better than B or

vice versa.

8.10. Type-1, Type-2 errors

Type-I errors are made when we reject a null hypothesis by marking a

difference significant, although no true difference exists. If we set up a low level of

significance the likelihood of type-I errors increase.

Type-II errors are made when we accept a null hypothesis by marking a

difference not significant, when a true difference actually exists. If we set up a high

level of significance the likelihood of type-II errors increase.

As a general rule it is probably wise to demand a significance level of at

least .01 in most experimental research, i.e., to risk type II errors by preventing those

of type I. But the .05 level is often satisfactory, especially in preliminary work.

8.11. Unit end questions

1. What is normal curve?

2. Discuss the important characteristics of a normal curve?
3. Area under the normal curve?
4. Deviations of the normal probability curve?
5. Explain Standard error, Significance, Null hypothesis, One tail test and two
tail test , Type-1, type-2 errors?

8.12. Source
1. Garrett, H.E. (1966) Statistics in Psychology and Education. Bombay : Vakils
Feefer & Simon Pvt. Ltd.
2. Guilford J.P. (1965). Fundamental Statistics in Psychology and Education
(4th Edn.). New Delhi: Subject Publications.
3. Mc.Guigan, F.J. (1990) Experimental Psychology New Delhi: Prentice Hall of
India Limited.
4. Singh A.K. (1997) Tests measurements and Research Methods in
Behavioural science Patna: Bharati Bhavan Publishers and Distributors.

112