THE t – TEST FOR

COMPARISONS OF TWO
INDEPENDENT GROUP
MEANS (BIVARIATE)
 The primary statistics used to determine
whether or not means from two different
samples are different beyond what would be
expected due to sample variation is called t –
test.

Formula:

1. For equal size

X1  X 2
t
2 2
S1 S2

N1 N 2
Degrees of Freedom

df = N1 + N2 – 2
N1 = number of scores, Group 1
N2 = number of scores, Group 2
X1 = mean of Group 1
X2 =mean of Group 2
S1 = variance of Group 1
S2 = variance of Group 2
2. For unequal size

X1  X 2
t
( N1  1) S1  ( N 2  1) S 2
2 2
 1 1 
  
N1  N 2  2  N1 N 2 
Steps for Computing the test
1. List the scores (x) for Group 1 in any order

2. Square each score (x2)

3. Add up the scores (X)

4. Add up the squared scores (X2)

5. Count the number of scores (N)

X
6. Compute the Mean: x 
N
7. Compute the variance: (S2)

8. Repeat steps 1 though 7 for Group 2

9. Compute t

10. Compare the obtained value of t to the

critical value of tc (if the absolute value of t
(ignoring the sign) is larger than the critical
value, reject the null hypothesis; if not,
accept the null hypothesis.
EQUAL SIZE EXAMPLE
Problem: How do the graduate students of Capitol
University compare in their performance in statistic
considering age?

Data: Group A Group B

(40 and below) (41and above)
3 14
7 18
10 19
16 20
18 20
Solution:

1. Null hypothesis

Ho: There is no significant difference in the

performance in statistics considering their age.

2. Alternative Hypothesis

H1: There is a significant; difference in the

performance in statistics considering their age.
3. Test Statistics

t < tc: Not Significant; Accept Ho

t > tc: Significant; Reject Ho

4. Rejection Region (at 0.05 level of significance)

df  N1  N 2  2  5  5  2  8
From table -------- (two tailed)

tc  2.306
5. Calculation of Test Statistics
 Group A Group B
( 40and below) (41 and below)

X1 X12 X2 X22
3 9 14 196
7 49 18 324
10 100 19 361
16 256 20 400
18 324 20 400

x1 =54 x12 = 738 x2 = 91 x22 = 1681

 X1  X 2
t
2 2
S1 S 2

N1 N 2

X 2 
 X 2

S2  N
N 1

738 
54 
2
1618 
91
2

S1 
2 5 and S 2 
2 5
5 1 5 1
 38.7  6.2
 54 91
mean; X 1   10.8 and X 2   18.2
5 5

hence;

10.8  18.2
t
38.7 6.2

5 5
t  2.48
6. Conclusion

Since the computed t is >tc, the null hypothesis is rejected,

therefore there is a significant difference in their
performance in statistics considering their age:

UNEQUAL SIZE EXAMPLE

Problem: How do the employees of Capitol University compare their job
satisfaction considering their positions?
Data: Group A Group B
(Faculty) (Deans and Heads)
16 20
9 5
4 1
23 16
19 2
10 4
5
2
Solution:

Hypothesis

1. Ho : There is no significant difference between

job satisfaction and their position.
2. H1 : There is a significant difference between
job satisfaction and their position.
3. Test Statistics:

t < tc; not significant: Accept Ho

t > tc; significant : Reject Ho
5. Rejection Region: at 0.05 level of significance

df = N1 + N2 – 2 = 8 + 6 = 12

from table: 0.05 and df = 12

tc = 2.179

6. Calculation of Statistics
 Group A Group B
(Faculty) (Deans and Heads)

X1 X12 X2 X22

16 256 20 400
9 81 5 25
4 16 1 1
23 529 16 256
19 361 2 4
10 100 4 16
5 25
2 4
∑ X1 =88 ∑ X12 =1372 ∑ X2 = 48 ∑X22 =702
 X1  X 2
t
N1  1S12  N 2  1S 2 2  1 1 
N1  N 2  2  N  N 
 1 2 

X 2 
X  2

S2  N
N 1

1372 
88
2
702 
48
2

S1 
2 8  57.7 and S 2 
2 6  63.6
8 1 6 1

X 88 48
Mean; X  hence; X 1   11 and X 2  8
N 8 6
 11  8
t
8  157.7   6  163.6  1  1 
 
862 8 6

t  0.72

7. Conclusion

Since the computed t < tc the null hypothesis is

accepted, therefore, there is no significant difference
of Capitol University employees job satisfaction and
their position.
Homogeneity of Variance
One important assumption behind the t –
test is that the variances of two samples
are similar, or homogenous. This is easily
checked by dividing the larger variance by
the smaller to compute on F as follows:
2

df  N L  1N s  1
SL
F 2
SS
where:

SL2 = the larger variance

SS2 = the smaller variance

df = degrees of freedom

NL = the number of scores in the group with the larger variance

NS = the number of scores in the group with the smaller

variance

Note: If the value (computed) is more than the value in the

table, you must conclude that the two variances are not
homogenous and you should use a non-parametric test
rather than t.
Example: A study comparing the physical
strength of learning-disabled boys to that of
non-disabled agemates. Each student’s
strength is tested in several ways and put on a
scale 1 to 10, the results are as follows:

Learning Disabled Learning Non – disabled

x  7.62 x  5.92
S  8 .4
2
S 2  5.12
N  13 N  31
Solution:

Test the homogeneity of variances

8.4
F  1.64
5.12
df  13  1, 31  1  12, 30
from table 0.05 level of significance

Fc  2.09
We cannot reject the null hypothesis that SL2 = SS2, thus the
variance can be considered homogenous and we can go
ahead and compute the t – test to compare the two means.

The t – test for comparisons of two means from matched

groups
(correlated sample)
Occasionally, we wish to test the difference between
the means of two matched groups for statistical
significance. This is most often done when we wish
to compare the same subjects under two different
conditions or at two different times. For example,
we might test a group of students on their
knowledge of statistics in November and then test
them again in March, to see if they have learned
anything at all. Or consider a situation in which
parents have complained that Engr. Cuanan gives
harder lessons in Physics than Engr. Ladica. The
Dean wants to find out if this is true, so he takes 20
students from a third teacher’s physics class and
gives them a 30 – item test made by Engr. Cuanan
and 30 – item test made by Engr. Ladica on the
same topic.
Formula for t with matched group:

D D
t or t 
ND  D 
2 2 2
SD
N N 1

D
Where : D  mean of the difference s scores 
N

D 2 
 D  2

S D  the var iance of difference scores 

2 N
N  10

N  np. of pairs of scores

Steps in Computing t for Matched Groups

1. Subtract each individual score on Group 2

from his/her score on Group 1 to get the
difference.

2. Add up the difference scores (D)

3. Compute the mean difference

4. Square each difference scores and add them

up (D)2
5. Compute the variance (SD2)

6. Compute the t

Example of t – test for correlated sample:

Problem: Is there a significant difference on

the results of the students in an open notes exam
and in close notes exam in statistics?
 Respondent Open Notes Close
Notes

1 36 30
2 50 55
3 43 41
4 60 50
5 53 40
6 36 36
7 51 42
329 294
Solution:

1. H0 : There is no significant difference on the result of the

student’s score in the open and close notes exam in
Statistics

2. H1 : There is a significant difference in the results of the

student’s score in the open and close notes exam in
Statistics.

3. Test Statistics:
t < tc : Not Significant : Accept H0
t > tc : Significant : Re ject H0

4. Rejection Region test at 0.05 level of significance

df = N – 1 = 7 – 1 = 6
tc = 1.943
5. Calculation of test statistics

Respondent Open Notes Close Notes D D2

13 6 30 6 36
2 50 55 -5 25
3 43 41 2 4
4 60 50 10 100
5 53 40 13 169
6 36 36 0 0
7 51 42 9 81

329 294 35 415

 D 35
t   2.09
ND  D  7415  35
2 2 2

N 1 7 1

D
or t 
S 2D
N

D 35
D  5
N 7

D 2 
D 
2
415 
35
2

SD 
2 N  7  40
N 1 7 1

5
t  2.09
40
7
6. Conclusion

The computed t value is greater than the critical tc value,

the null hypothesis is rejected therefore, there is a
significant difference on the result of the students in the
open and closed notes exam in Statistics.