Probability and Statistics Unit 13

Unit 13 Tests of Significance Based on t,

F and Z Distribution
Structure:
13.1 Introduction
Objectives
13.2 t- test Distribution
Properties of students t- distribution
Applications of t- distribution
13.3 F Test Distribution
13.4 Z- Test Distribution
13.5 Summary
13.6 Terminal Questions
13.7 Answers

13.1 Introduction
In the previous unit we discussed about various tests of hypothesis and test
of significance based on large samples. The assumptions taken for large
sample test does not hold good for small sample tests. In this unit we will
study about small sample tests. There are various tests for small samples.
Here in this unit we will study t-test, F-test and Fishers Z- test.
Objectives:
At the end of this unit the student should be able to:
Define student t test, Fishers Z test and F test
Explain the important properties and applications of these tests

13.2 t- Test Distribution

It is known as students t- test, obtained by W.S. Gosset in the year 1908.
This is small sample test.
Let be a random variable of size n from a normal population
with mean and variance . Then the students t is defined by the statistic

Sikkim Manipal University Page No.: 378

Probability and Statistics Unit 13

Where, , is the sample mean and is an

unbiased estimate of of the population variance and it follows the
Students t- distribution with (n-1) degrees of freedom

13.2.1 Properties of Students t- distribution

1. The t- distribution like the standard normal distribution is bell- shaped
and symmetrical around mean zero.
2. The t- distribution ranges from to .
3. The shape of the t- distribution changes as the number of degrees of
freedom changes.
4. The variance of the t-distribution is always greater than one and is
defined only when degree of freedom and is given as

5. The t- distribution is more platykurtic (less peaked at the centre and

higher in tails) than the normal distribution.
6. The t- distribution has a greater dispersion than the standard normal
distribution. As n gets larger, the t- distribution approaches the normal
form. When n is large as 30, the difference is very small.
Note: The calculated value of t is compared with the tabulated value of t for
a given degrees of freedom at different specified level of significance.
We establish a null hypothesis and if the calculated t value is greater than
the tabulated value of t, we reject the null hypothesis. If the calculated value
of t is smaller than the tabulated value of t, we accept the null hypothesis.

13.2.2 Application of t- distribution

1. Test of Hypothesis about the population mean:
When the population distribution is normal and standard deviation is
unknown then the t statistic is defined as:

follows the t distribution with (n-1) d.f.

where = sample mean, hypothesized population mean, n = sample
size and s is the standard deviation of the sample calculated by the formula:

Sikkim Manipal University Page No.: 379

Probability and Statistics Unit 13

And the null hypothesis in this case is whether there is significant difference
between .
Example: Ten oil cans are taken at random from an automatic filling
machine. The mean weight of the cans is 15.8 kg and standard deviation is
0.50 Kg. Does the sample mean differ significantly form the intended weight
of 16 kg?
Solution: Null hypothesis: Sample mean weight is not different from the
intended weight.
Here
By t- test, we have

Here and the tabulated value of t for 9 d.f. at 5% level of

significance is 2.26and the computed value of t is smaller than the tabulated
value. Therefore, the difference is insignificance and the null hypothesis is
accepted. Hence the difference between sample mean weight and the
intended weight is insignificant.
Example: A group of 5 patients treated with medicine A weighs: 42, 39, 48,
60 and 41 kg. In the light of above data discuss the situation that mean
weight of the population is 48 kg. Test at 5 % level of significance. Given
that tabulated value of t at 4 d.f is 2.776.
Solution: Let the null hypothesis is mean weight of the population is 48 kg.
Weight X (
42 -4 16
39 -7 49
48 2 4
60 14 196
41 -5 25

Sikkim Manipal University Page No.: 380

Probability and Statistics Unit 13

Applying t- test, we have

Since for 4 d.f. the calculated value of t is less than the tabulated value of t ,
we accept the null hypothesis and therefore we conclude than the mean
weight of population is 48 kg.
SAQ 1: A random sample of size of 9 boys had heights (in inches): 45, 47,
50, 52, 48, 47, 49, 53 and 51. In the light of the data, discuss the suggestion
that the mean height in the population is 47.5. Given that the tabulated value
of t for 8 d.f at 5% level = 2.306.
SAQ 2: Prices of shares (in Rs) of a company on the different days in a
month were found to be: 66, 65, 69, 70, 63, 64 and 68. Test whether the
mean price of the shares in the month is 65.
2. Test of Hypothesis about the difference between two means:
Let two independent random samples of sizes and taken from normal
population with means and and standard deviation . So to
test that the two population means are equal or the difference is
significant, we use t- test and the statistic is given by

And s can be computed as

Example: In a test given to two group of students, marks obtained are as

follows:

Ist Group 18 20 36 50 49 36 34 49 41
IInd Group 29 28 26 35 30 44 46

Sikkim Manipal University Page No.: 381

Probability and Statistics Unit 13

Examine the significance of difference between the mean marks secured by

the students of the above two groups. (The value of t at 5% level for 4 d.f =
2.14)

Solution: Null Hypothesis: Let there is no significance difference in the

mean marks of the two group of students, i.e.,

2 2
Group I(X1) (X1- ) Group II(X2) (X2- )
(X1- ) (X2- )
18 -19 361 29 -5 25
20 -17 289 28 -6 36
36 -1 1 26 -8 64
50 13 169 35 1 1
49 12 144 30 -4 16
36 -1 1 44 10 100
34 -3 9 46 12 144
49 12 144
41 4 16

t- test:

Degrees of freedom

For for 2 tailed test is 2.14. Since the calculated value of t is

less than the table value, we accept the null hypothesis and conclude that
the mean marks of the students of 2 groups do not differ significantly.

Sikkim Manipal University Page No.: 382

Probability and Statistics Unit 13

Example: Two independent samples of 8 & 7 items gave the following

values:
Sample X1 9 11 13 11 15 9 12 14
Sample X2 10 12 10 14 9 8 10

Examine whether the difference between the means of two samples is

significant at 5% level. (The value of t at 5% level for 13 d.f = 2.16)

Solution: Null Hypothesis: Let there is no significance difference in the two

mean

2 2
Group (X1) (X1- Group (X2) (X2-B)
(X1- (X2-
9 -3 9 10 0 0
11 -1 1 12 2 4
13 1 1 10 0 0
11 -1 1 14 4 16
15 3 9 9 -1 1
9 -3 9 8 -2 4
12 0 0 10 0 0
14 2 4

Since the actual means are not whole numbers, we take 12 as assumed for
X1 and 10 as assumed for X2

Applying t- test:

Degrees of freedom

Sikkim Manipal University Page No.: 383

Probability and Statistics Unit 13

For for 2 tailed test is 2.16. Since the calculated value of t is

less than the table value, we accept the null hypothesis and conclude that
the mean marks do not differ significantly.
SAQ3: Two salesman A and B are working in a certain district. From a
sample survey conducted by Head Office, the following results were
obtained. State whether there is any significant difference in the average
sales between the two salesmen.

A B
No. of sales 10 18
Average sales (in lakh Rs.) 190 205
Standard Deviation (in lakh Rs) 20 25

3. Test of Hypothesis about the difference between two means with

dependent samples i.e paired data
When the sample items are the same but different situations are be
analyzed then in this case the test of significance of the difference between
two means in two situations, the appropriate test statistics t to be used is

Where is mean of difference i.e., , n = size of the sample, s = standard

deviation

If the computed value of t is less than the table value of t, null hypothesis is
accepted, otherwise rejected.
Example: Ten persons were appointed in an officer cadre in an office. Their
performance was noted by giving a test and the marks were recorded out of
100. They were given 4 months training and a test was held and marks were
recorded out of 100.

Sikkim Manipal University Page No.: 384

Probability and Statistics Unit 13

Employee : A B C D E F G H I J

Before

Training: 80 76 92 60 70 56 74 56 70 56

After

Training: 84 70 96 80 70 52 84 72 72 50

By applying the t- test, can it be concluded that the employees have

benefited by the training?
Solution: Null hypothesis: Employee have not been benefited by the
training.
st nd st nd 2
Employee Before 1 After 2 (1 2 )d d

A 80 84 -4 16

B 76 70 6 36

C 92 96 -4 16

D 60 80 -20 400

E 70 70 0 0

F 56 52 4 16

G 74 84 -10 100

H 56 72 -16 256

I 70 72 -2 4

J 56 50 6 36

n=10

where

D.f , For

Sikkim Manipal University Page No.: 385

Probability and Statistics Unit 13

The calculated value of t is less than the table value, so the null hypothesis
is accepted.
SAQ 4: Ten workers were given a training programme with a view to
shorten their assembly time for a certain mechanism. The results of the time
and motion studies before and after training programme are given below:

Worker : 1 2 3 4 5 6 7 8 9 10
st
1 study: 15 18 20 17 16 14 21 19 13 22
nd
2 study: 14 16 21 10 15 18 19 16 14 20

On the basis of this data can it be concluded that the training programme
has shortened the average assembly time.
4. Test of Hypothesis about correlation coefficient:
Suppose that a random sample of size n has been drawn from a
bivariate normal population and let r be the observed sample correlation
coefficient. To test the hypothesis that the correlation coefficient in the
population is zero i.e., or the observed correlation is significant or not,
we use t-test and the appropriate test statistic t to be used is

Where r = sample correlation coefficient, n = size of sample

Example: A correlation coefficient of 0.6 is discovered in a sample of 18
observation. Is it significant at 1% level?
Solution: Given, r = 0.6, n = 18
(Two tailed test)
Applying t- test

Degree of freedom, v = n-2 = 18-2 = 16

For v = 16, t0.05 for two tailed test = 2.92
Since the calculated value of |t|= 3 is greater than the table value of t, we
reject H0 and concluded that the observed valued of the correlation
coefficient is significant.

Sikkim Manipal University Page No.: 386

Probability and Statistics Unit 13

Example: A random sample of 27 observations from a normal population

gives a correlation coefficient of -0.4. Is this significant of the existence
correlation in the population? (Given for v = 25, t0.01= 2.79)
Solution: We are given : n = 27, r = -0.4 and
Applying t- test

Degree of freedom, v = n-1 = 27- 2 = 25

For v = 25, t0.05 for two tailed test = 2.79
Since the calculated value of |t| is less than the table value of t, we accept
H0 and concluded that the correlation of the sample is not significant to
warrant the existence of such correlation in the population.

13.3 F- Test Distribution

F- distribution is also known as Variance Ratio distribution. It is defined in
terms of ratio of the variances of two normally distributed populations. It is

with degrees of freedom, where

is the unbiased estimator of and is the

unbiased estimator of .

If , then the statistic follows F-distribution with

degrees of freedom.
Example: Two sources of raw materials are under consideration by a
company. Both sources seem to have similar characteristics but the
company is not sure about their respective uniformity. A sample of 10 lots
from A yields a variance of 225 and a sample of 11 lots from source B yields
a variance of 200. Is it likely that the variance of A source is significantly
greater than the variance of source B at 1% confidence interval.

Sikkim Manipal University Page No.: 387

Probability and Statistics Unit 13

Solution: Null hypothesis: , i.e., the variance of source A and

that of source B are same. The F statistics to be used here is

Where ,

The table value of F for at 1% level of significance is

4.49. Since the computed value of F is smaller than the table value of F, the
null hypothesis is accepted. Hence, the population variances of the two
populations are same.
Example: Two random samples drawn from normal populations are:

Sample I: 20 16 26 27 23 22 18 24 25 19

Sample II: 27 33 42 35 32 34 38 28 41 43 30 37

Obtain estimates of the variances of the population and test whether two
populations have same variance.
Solution: Null Hypothesis ; two populations have the same variance
x1 Sample I x2 Sample II

20 -2 4 27 -8 64
16 -6 36 33 -2 4
26 4 16 42 7 49
27 5 25 35 0 0
23 1 1 32 -3 9
22 0 0 34 -1 1
18 -4 16 38 3 9
24 2 4 28 -7 49
25 3 9 41 6 36
19 -3 9 43 8 64
30 -5 25
37 2 4

Sikkim Manipal University Page No.: 388

Probability and Statistics Unit 13

Since the numerator is greater then denominator, therefore,

SAQ5: We want to test whether male and female students have different
variances in their test scores on a certain course. The 25 male students
have a sample variance of and 31 females students have a
sample variance of test the hypothesis that the variances
are equal at the 5% level of significance, using two tailed test.
SAQ6: The table below represents the time taken to do a job by method I
and method II by workers. Can we conclude that the variance of time
distribution for method I and method II are same?

Method I 27 23 16 20 26 22
Method II 33 35 34 27 42 32 38

13.4. Z- Test Distribution

If the population correlation coefficient is other than zero than zero or
difference between two sample correlation coefficients are to be tested, then
the t- test can not be used. Fishers Z test has two applications:
1. To test whether an observed value of r differs significantly from same
hypothetical value of population correlation coefficient, other than zero,
the appropriate test statistic to be used is

|Z|

Where and

Note: If is not known, then it is taken as zero in which case

|Z|

Sikkim Manipal University Page No.: 389

Probability and Statistics Unit 13

The calculated value of Z is compared with 1.96 at 5% level of significance

and 2.58 at 1 % level of significance.
If |Z| > 1.96, the difference is considered significant at 5 % level of
significance otherwise insignificant.
If |Z| > 2.58, the difference is considered significant at 1 % level of
significance otherwise insignificant.
Example: Test the significance of the correlation of coefficient r = 0.5
discovered in a sample of 19 paired of observation against hypothetical
correlation . apply Fishers Z test.
Solution: Let us take the hypothesis that correlation coefficient in the
population is 0.7 i.e.,

Applying z transformations, we obtain

|Z|

Since the calculated value of |Z| is less than 1.96, we accept the null
hypothesis at 5% level and concluded that the population correlation
coefficient is 0.7.
2. Testing the significance of the difference between two independent
sample correlation coefficients.
The Z transform for two samples is

Sikkim Manipal University Page No.: 390

Probability and Statistics Unit 13

And standard Error is

With this we can compute |Z| as follows

|Z|

If the calculated value of |Z| is greater than 1.96 at 5% level of significance

than the difference between 2 rs is significant.
Example: The following datat give sample sizes and correlation coefficient .
test the significance of the difference between the values using Fishrs- Z
test
Sample Size Value of r

5 0.87

12 0.56

Solution: Let us take the hypothesis that two correlation coefficients do not
differ significantly i.e., and

|Z|

As the value of |Z| is less than the 1.96 at 5% level of significance so we

accept the null hypothesis and conclude that the difference is not significant.

13.5 Summary
In this unit we studied about different test for small samples and solved
several problems based on these tests. Practical applications of tests are also
discussed.
Sikkim Manipal University Page No.: 391
Probability and Statistics Unit 13

13.6 Terminal Questions

1. A soft drink vending machine is set to dispense 8 ounces per cup. If the
machine is tested 9 times yielding a mean cup fill of 8.2 ounces with a
standard deviation of 0.3 ounce, what can we conclude about the null
hypothesis of ounces against the alternative hypothesis of
ounces at .
2. A random sample of size 10 has mean as 40 and standard deviation 5.
Can this sample be regarded as taken from the population having 42 as
mean. (For v = 9,
3. The mean life of sa,ple of 10 electric light bulbs was found to be 1456
hrs with s = 423 hrs. A second sample of 17 bulbs chosen from a
different batch showed a mean life of 1280 hrs with s = 398 hrs. Is there
a significant difference between the means of the two batches. Here s is
the standard deviation.
4. A certain stimulus administered to each of 12 patients resulted in the
following increase in blood pressure: 5,12,8,-1,3,0,-2,1,5,0,4,6. Can it be
concluded that the stimulus will in general be accompanied by an
increase in blood pressure. (for v = 11, t0.01=2.21)
5. The sales data of an item in six shops before and after a special
promotional campaign are as under;

Shops: A B C D E F
Before
Campaign: 53 28 31 48 50 42
After
Campaign: 58 29 30 55 56 45

Can the campaign be judged to be a success? Test at 5% level of

significance.
6. A random sample of 25 from a normal universe gives correlation
coefficient of -0.48. Is this significant of the existence of correlation in the
population?

Sikkim Manipal University Page No.: 392

Probability and Statistics Unit 13

7. A random sample of 27 pairs of observation from a normal population

gave a correlation coefficient of 0.58. Is this significant of correlation in
the population.
8. Tets the significance of the correlation r = 0.75 from a sample of size 30
against hypothetical correlation
9. From a sample of 10 pairs of observation the correlation is 0.5 and the
corresponding population value is 0.3. Is this difference significant at 5%
level of significance?

13.7 Answers
Self Assessment Questions
1. Null hypothesis: Mean height in the population is 47.5
Height X A=49, d = (X-A)
45 -4 16
47 -2 4
50 1 1
52 3 9
48 -1 1
47 -2 4
49 0 0
53 4 16
51 2 4

Since the actual mean is in fraction, we should take deviation from

assumed mean 49

Sikkim Manipal University Page No.: 393

Probability and Statistics Unit 13

Applying t- test, we have

Since the calculated value of t is less than the tabulated value of t, so we

accept the null hypothesis.
2. Null hypothesis: The mean price of the shares in the month is not 65.
Assuming the population to be normally distributed and the population
standard deviation is unknown .
X A=65, d = (X-A)

66 1 1
65 0 0
69 14 16
70 35 25
69 4 16
71 6 36
70 5 25
63 -2 4
64 -1 1
68 3 9

Applying t- test, we have

Sikkim Manipal University Page No.: 394

Probability and Statistics Unit 13

The tabulated value of t for 9 d.f at 5% level of significance is 2.2.

Since the computed value of t = 2.81 is greater than the table value, we
reject the null hypothesis. Hence, the mean price of the shares in the
month is not 65.
3. Null Hypothesis: there is no significant difference in the average sales
between the two salesmen.

Applying t test:

Where

4. Null hypothesis: Training programme has not helped in reducing the

average assembly time.
st nd st nd 2
Worker 1 Study 2 Study (1 2 )d d
1 15 14 1 1
2 18 16 2 4
3 20 21 -1 1
4 17 10 7 49
5 16 15 1 1
6 14 18 -4 16
7 21 19 2 4
8 19 16 3 9
9 13 14 -1 1
10 22 20 2 4

n=10

where

D.f , For

Sikkim Manipal University Page No.: 395

Probability and Statistics Unit 13

The calculated value of t is less than the table value, so the null
hypothesis is accepted.
5. Null Hypothesis is

Lets first calculate and . We have ,

so

Therefore,

d.f in numerator = 25-1= 24 and d.f. in denominator is 30-1= 29. Since

the calculated value of F at this d.f is 1.87 is less than the tabulated
value of 2.154, we can not reject the null hypothesis. So we do not have
sufficient evidence to conclude that male and female students have
sufficient variances.
6. The tables A and B represent the frequency table required for the
calculation of sample means for the data given for two different
methods.
Table A Required values of the method I to calculate sample mean
X d = X - 22 d2
27 5 25
23 1 1
16 -6 36
20 -2 4
26 4 16
22 0 0
Total 2 82

Sikkim Manipal University Page No.: 396

Probability and Statistics Unit 13

Table B Required values of the method II to calculate sample mean

2
X d = X 35 d
33 2 4
35 0 0
34 1 1
27 8 64
42 7 49
32 3 9
38 3 9
Total 4 136

1
2
21

S

d2(

d
)
824
/
6
16
.
266
1
n
1 n5
1 1

1 2(
d
)2
2
S
2 n1



d
n =
1
136 16 / 7= 22.286
2 2 6
2 2
Null hypothesis Ho: 1 = 2 , that is, the sample variances of two
methods are equal.
2 2
Alternate hypothesis H1: 1 2 , that is, the null hypothesis is
wrong.
Level of significance 5 % and D.O.F , from the F-table:
Ftab = 4.95

S 22
Test Statistics: F = 2
S1

22.
2286
F 1
.37
16
cal .266
Conclusion: Since Fcal (1.37) < Ftab (4.95), H0 is accepted. Hence, there
is no significant difference.

Sikkim Manipal University Page No.: 397

Probability and Statistics Unit 13

Terminal Questions
1. t = 1.885, accept Null hypothesis.
2. t = 1.2, accept Null hypothesis
3. s- statistics = 407.17, t = 1.0806
4. t = 2.90, reject Null hypothesis
5. t = 2.78, reject Null hypothesis
6. t = 2.624, reject Null hypothesis
7. t = 3.56, reject Null hypothesis
8. |Z|= 1.8, Accept Null hypothesis
9. |Z|=0.96, Accept Null Hypothesis

__________________

Sikkim Manipal University Page No.: 398