The Student's t-test, also known as the t-test, is a statistical test used to compare the means of

two groups. It's particularly useful when the population standard deviation is unknown,
which is often the case when dealing with small samples. The t-test helps determine if the
observed difference between the means is statistically significant or likely due to random
chance.

Applications:
Comparing means of two groups: The t-test is used to compare the means of two unrelated
groups (e.g., men vs. women, treated vs. control).
Testing a hypothesis about a population mean: The t-test can be used to determine if a sample
mean significantly differs from a hypothesized population mean.
Paired samples: It can also be used to compare the means of two related groups, where each
subject is measured twice (e.g., before and after a treatment).

Merits:
Useful for small samples: The t-test is robust and can be used even with relatively small sample
sizes (less than 30 observations).
Flexible in assumptions: While the t-test assumes normality of the population, it is relatively
robust to deviations from normality, especially with moderate sample sizes.
Wide applicability: The t-test has numerous applications in various fields, including medicine,
psychology, and social sciences.
Demerits:
Assumes normality: While robust, the t-test's accuracy can be affected if the population is not
normally distributed, especially with small sample sizes.
Does not account for multiple comparisons: If conducting multiple t-tests, the risk of finding a
false positive (Type I error) increases.
Can be less sensitive than other tests: In some cases, other statistical tests may be more
sensitive to detecting differences between groups.
STATISTICAL INFERENCETESTS OF HYPOTHESES 911

interested not so much in the value

lation. but more generally whether this ofvalue
the correlation in the
parenl
could
ncorrelated population, Le. whether it is significant of have arisen from an
parent population. correlation in the
It shoulid be noted that the investigator who works with
et know that his estimates will vary widely from very small samples
Eer he must be satistied with relatively wide sample to sample.
Drecision of statement 1S less, O1 cOurSe, the wider theconfidence
intervals
intervals.
emploved.
Each inference drawn from large sample results in far more precise in the
limits it sets up than is an inference based on a much smaller sample.

The Assumption of Normality

While dealing with small samples also, an assumption is made that the
parent population is normal, unless otherwise stated. Strictly speaking.
therefore, our results will be true only for the normal population. However.
as pointed out earlier, the assumption of normality is not very much
warranted in case of small samples. Experiments have, therefore, been
made to ascertain whether the results are true for other types of population.
Theoretical work confirms that the results remain true for populations
which do not deviate markedly from normality. However, if there is any
good reason to suspect that the parent population is markediy skew. Le.
æ, or J-shaped, the methods given below cannot be applied with much
confidence.
Since in many of the problems it becomes necessary to take a small size
sample, considerable attention has been paid in developing suitable tests for
dealing with problems of small samples. The greatest contribution to the
theory of small samples is that of Sir William Gosset and R.A. Fasher. Sir
William Gossetpublished his discovery in 1905 under the pen name Student.
He gave a test popularly known as 't-test' and Fisher gave another test known
as 'z-test'.These tests are based on 'e-distribution and 'z-distribution.
Student's t-Distribution

Theoretical work on t-distribution was done by W.S. Gosset (1876-1937) in

the early 1900. Gosset was employed by the Guinness & Son, a Dublin
bravery. Ireland, which did not permit employees to publish research findings
under their own names. So Gosset adopted the pen name "student and
published his findings under this name. Thereafter, the t-distribution is
commonly called Student's tdistribution or simply Student's distribution.
The t-distribution is used when sample size is 30 or less and the population
standarddeviation is unknown.
The "t-statistic" is defined as :

where S= V2x-X²
n-1
The t-distribution has been derived mathematically under the assumption of
a normally distributed population.
 912 STATISTICAL METHODS

It has the following form 1)/2

where t= (X- 9) Vn
S
C=a constant required to make the area under the curve equal to unity
v=n- 1, the number of degrees of freedom.

Properties of t-Distribution

1. The variable t-distribution ranges from mínus infinity to plus

infinity
2. The constant c is actually a function of v (pronounced as nu), s0
that for a particular value of v, the distribution of f(t) is completely
specified. Thus f() is a family of functions, one for each value of y.
3. Like the standard normal distribution, the t-distribution is
and has a mnean zero symmetrical
4. The variance of the
t-distribution is greater than one, but
approaches one as the number of degrees of freedom and, therefore.
the sanple size becomes large. Thus the variance of the t-distribution
approaches the variance of the standard normal distribution as the
sample size increases. It can be demonstrated that from an infinite
number of degrees of freedom (v= o), the t-distribution and normal
distribution are exactly equal. Hence there is a widely practised
of thumb that samples of size n> 30 may be considered large rule
the standard normal distribution may appropriately be and
used as an
approximation to t-distribution, where the latter is the theoretically
correct functíonal form.
The following diagram compares one normal distribution with two t-distributions
of different sample sizes :

Normal Distribution
t-Distribution t-Distribution
for Sanple Size for Sample Size
n=15 n=2

Normal Distribution, t-Distribution for Sample Size n=15, and

-Distribution for Sample Size -2.
The above diagram shows two important characteristics of t-distribution.
First, a tdistribution is lower at the mean and higher at the tails than a
normal distribution. Second, the t-distribution has proportionately greater
area in its tails than the normal distribution. Interval widths from tdistributions
are, therefore, wider than those based on the normal distribution.
 STATISTICAL INFERENCE TESTS OF HYPOTHESES
913

rhe tTable. The t table given at the end is the probability integral of
distribution. It gives, over a range of values of v. the probabilities of
exceeding by chance value of t at different levels of significance. The
t-distribution has a different value for each degree of freedom and when
degrees of freedom are infinitely large. the tdistribution is equivalent to
rmal distribution and the probabilities shown in the normal distribution
tables are applicable.

Application of the t-Distribution

The following are sonie of the examples to illustrate the way in which the
Student' distribution is generally used to test the significance of the
various results obtained from small samples.

1. To Test the Significance of the Mean of a Random Sample In

determining whether the mean of a sample drawn from a normal
population deviates significantly from a stated value (the hypothetical value
of the populations mean), when variance of the population is unknown we
calculate the statistic :
t=
(X-) Vn
S
whereOX= the mean of the sample
u=the actual or hypothetical mean of the population
n=the sample size
S= the standard deviation of the sample
or S=
n-1 n-1

(n- 1)
whered= deviation from the assumed mean.
we say that the difference
If the calculated value of t exceeds to.05:
exceeds to.o, the difference is
between X and u is significant at 5% level, if it
we conclude that the difference
said to be significant at 1% level. If t| < to 05, sample might have been
between X and u is not significant and hence the
drawn from a population with mean = .
that the sample is a random
Flducial Limits of Population Mean. Assuming
unknown mean the 95% tiducial
Sample from a normal population of
limits of the population mean (4) are :
S
X Jn lo o5
and 99% limits are

XtJ o 05
The following examples will illustrate this test : that his bulbs
make of electric bulbs claims
llustration 28.. The manufacturer of a certain deviation
months with a standard of 5 months. A random sample of
e a mean life of 25
6 such bulbs gave the following values.
 914 STATISTICAL METHODS

Life of months 24, 26, 30, 20, 20, 18,

Can you rogard the producor's claim to be valid at 1% level of significance ? (Givon that
table values of the appropriate test statistics at the said lovel are 4032, 3.707 and 499 the
and 7 degrees of freedom respectively.) for 5,6
Solution. Lot us take the hypothesis that thera is no significant differonco in the
of bulbs in thesample and that of the population. Applying tost : moan \lfo
S
CALCULATION OF X and S
X (X- X) 2
X

24 1
26 +3ad 9
30 +7 49a
20 -3 9
20
18 -5 25
EX= 138
Ex= 102
X- X 138 23
6

-V102
n-1 5
V20.4 =4517
23- 25|2x2449 1-084
4517 4517
V=n-1 =6-1= 5. For v=5. to.01 = 4-032.
The calculated value of t is less than the table value.
producers claim is not valid at 1% level of significance. The hypothesis is accepted. Hence, the
Illustration 29. A random sample of size 16 has 53 as mean. The
the deviations taken from mean is 135. Can this sample be sum of the squares of
regarded
population having 56 as mean ? Obtain 95% and 99% confidence limits ofas the taken from the
population. (for v= 15, to o5 =213 for V= 15, to01 =295) mean of the
(M.Com., Madras Uni., 1998; M.Com., Annamalat Univ., 2003: CAFC,
2006)
Solution. Let us take the hypothesis that there is no
sample mean and hypothetical population mean. significant difference between the
Applying ttest :

S
X=53, u =56, n=16, X(X- X)= 135
s- V2(X-
n-1
X)V135 =3
15
t 53-56| V16 3x4 4
3
V 161 15. For v=16, to.05
The calculated value of tis more than the =2:13
sample has not come from a population table value, The hypothesis is rejected. Hence, the
having 56 as mean.
 eNNTISICAL INFERENCE-TESTS OFIYTOESES 915

QH% ntoerxe inits of the popuation mean

83*
V16ex213
-53 16 =514 to 54.6
Qo9% ntcernve limits of the population mean
Vn
-63 t x295
53+ 2 . 9 5 e u t
4
53 +2212 =50788 to 55212
lustration30. The life time of electric bultbs for a random sample of 10 from a large
consignment gave the following data
Item 1 2 4 5 6 7 8 9 10
Life in '000 hours 42 46 39 41 62 38 39 43 4.4 56
Can we accept the hypothesis that the average life time of bulbs is 4,000 hours.
(M.Com., Madras Univ., 2005)
Solution. Let us take the hypothesis that there is no significant difference in the sample
mean and the hypothetical population mean. Applying the test.

CALCULATIONOF Xand S
(X (X-2
42 -02 ob004
46 6bnkehe + 0 2 b 0.04
39
41 -03 0-09

52 +0-8 0-64

38 -06 036
39 t -05al 025
43 -01 ie 001

44 0 0

56 +1-2 144

XX= 44 E(X- 0 =312

V312 0.589
S= n-1 9
44- 4 V10 - 04x3162 2 148
0589 0589
V=n-1 = 10- 1 = 9, For v 9, to 05 =2262
accepted. The
e calculated value of t is less than the table value. The hypothesis is
average life time of the bulbs could be 4000 hours.