The choice of whether to use a non directional

test depends upon several factors .If the nature of

the study is such that there is very little basis for
guessing the outcome of the study, then the non-
directional test is generally chosen.
In the course of comparing one group with
another and some indicators are available to favor
one over the other, then a directional test may be
used.

The procedure in hypothesis testing can be
summarize in the following steps;
Step 1; Formulate the null hypothesis (Ho) that
there is no significant difference between
items being compared. State the
alternative hypothesis(Ha) which is used in
case Ho is rejected.
Step 2; Set level of significant,
Step 3; Determine the test to be used. Used z-test
if population standard deviation is given,
and t-test if the standard deviation is
given from the sample.

Step 4; Determine the tabular value for the
test. For a z-test, used the table of the
critical values of z based on the area of
the normal curve. For a t-test, one must
first compare for the degrees of
freedom, then look for the tabular value
from the table of t-distribution. For a
single sample, df= number of items -1
or in symbols. df= n-1


For two samples, the degrees of
freedom is calculated as df= n+n-2
Where n refers to the number of items
in the first sample and n refers to the
number of items in the second sample.
Step 5; Compute for Z or T as needed,
using any of the following
formula;

X
Z
N

= ( )
X N
Z

o

=
Where:
= sample mean
= population mean
= population standard derivation
n = no. of items in the sample
2. Comparing two sample means




Where:

X

o
1
2
1
2
mean of the first sample
= mean of the second sample
n number of items in the first sample
n number of items in second sample
population standrad
X
X
deviation o
=
=
=
=
1
1 2
2
or
1 1
X X
z
n n
o

=
+
1 2
2 2
1 2
1 2
X X
Z
s s
n n

=
+
The t-test is a statistical tool we can use in analyzing
small samples, usually less than thirty, without having to
worry about distorted result.
1.Sample mean compared with population
mean
( ) 1 x n
t
s

=
1
x
t
s
n

=

Where;
= sample mean
= population mean
s = sample mean
n= number of items in the sample
X

2. Comparing two samples for

independent samples





Where;
x= mean of the first sample
x= mean of the second sample
S= Standard deviation of the first
S= standard deviation of the second
n= number of items in the first sample
n= number of items in the second sample

( )( ) ( )( )
1 2
2 2
1 1 2 2
1 2 1 2
1 1
1 1

2
X X
t
n S n S
n n n n

=
(
+ (
+
(
(
+

(

1 2
2 2
1 2
1 2
X X
t
s s
n n

=
+


Where;
= average of the differences between the given
items, the pre-test and the post-test



= Standard deviation of (x-x)
n= Number of pairs in the sample

2( )2
1
d
t
n d d
n
=

d
t
n
o
=
d
1 2
1
( )
n
n
x x
d
n

=

Step 6; compare the computed value
with its corresponding tabular value
then state your conclusion based on
the following guidelines.
A.Reject ho,
If the absolute computed value is
equal to or greater than the
absolute tabular value. There is a
significant differences.
B. Accept ho,
If the absolute computed value is
less than the absolute tabular value.

A. Sample mean compared with population
A researcher knows that the average height of Filipino
women is 1.525 meters. A random sample of 26
women was Taken and was found to have a mean
height of 1.56 meters, with a standard deviation of 0.10
meter. Is there a reason to believe that the 26 women in
the sample are significantly taller than the others at 0.05
significant level?
Step 1; Ho, the sample is not significantly taller than the
other Filipino women.
Ha, the sample is significantly taller than the
others
Step 2; Set = 0.05
Step 3; The standard deviation given is based on the
sample so t-test must be used.


( =1.525) or Ho: x x =

( >1.525) or Ho: x x >
Step 4. Look for the degree of freedom. Since only the sample is
given, df=n-1
df =26-1
df=25
Step 5. The given values in the problem are:








Step 6. The absolute value is greater than the absolute tabular value.
The Ho is rejected (The Samples is significantly taller than the
others.)
X=1.56 meters = 1.525 meters
26 s = 0.10
X- 0.035 5
t= t= =
s
1 0.10
n-1
1.56-1.525
t= t= 1.75
0.10
0.353
t=
0.10
25
n

=
| | | |
| |
\ . \ .
B. Comparing Two Sample Means or Independent
Samples
A teacher wishes to test whether or not the modular method of
teaching is more effective than the traditional method. She picks two
classes of approximately equal intelligence ( verified through an
administered IQ test). She gathers a sample of 18 students to whom she
uses the modular method and another sample of 14 students to whom
she uses the traditional method. After the experiment, an objective test
revealed that the first sample got a mean score of 28.6 with a standard
deviation of 5.9, while a second group get a mean score of 21.7 with a
standard deviation of 4.6. Based in the result of the administered test, can
we say that the modular method is more effective than the traditional
method.
Step 1; Ho; The modular method is as effective as the traditional method.


Ha; The modular method is more effective than the traditional
method.


1 2
; Ho x x =
1 2
Ha x x = >

Step 2;
Set = 0.05
Step 3;
Since the standard deviations or the samples are given,
the t-test is used.
NOTE; x

= modular method
x

= traditional method
Step 4
; Look for the degrees of freedom. For two samples
df= n

+n

-2Z
= 18 + 14 -2
=30
Step 5
. The given values in the problem are;

X

= 28.6 S

= 5.9 n

= 18
X

= 21.7 S

= 4.6 n

= 14


( )( ) ( )( )
( )( ) ( )( )
( ) ( )
| |
1 2
2
1 1 2 2
1 2 1 2
2
1 1
1 1
2
28.6 21.7
18 1 5.9 14 1 4.6
1 1
18 14 2 18 14
6.9

17 34.81 21.16
0.06 0.07
32 2
6.9
t =
(28.895)(0.13)
6.9

3.756
6.9
t=
1.94
t=3.56
X X
t
n s n s
n n n n
t
t
t

=
(
+ (
+ (
(
+
(

=
(
+
(
+ (
(
+

(

=
+ (
+
(

=
Step 6; The computed value is in the
rejection region. It is greater than the
tabular value 1.697. the Ho is rejected.
(the modular Method of teaching is
more effective than the traditional
method.

A group of employees was given in attitude test
on E-VAT. Then, they were shown a film favorable
to implementation of E-VAT. The test was then re-
administered. Is there a significant deference bet.
Their attitude before and after they have seen the
film? Use = .05. the data are shown below.
x x
15 25
17 19
19 23
23 27
23 30
21 20
19 23
18 21
9 17
8 17
0 1 2
1.1 : H =
1: 1 2
1.2 H =
1: 1 2
1.2 H =
1.3 interval measurement
1.3 interval measurement
1.5 =.05 n=10 o
x X D
15 25 10 100
17 19 2 4
19 23 4 16
23 27 4 16
23 30 7 49
21 20 -1 1
19 23 4 16
18 21 3 9
9 17 8 64
8 17 9 81
2 2
( ) D X X
1
172 X =

2
222 X =

50 D =

2
356 D =

( )
2
2
1
D
t
N D D
N
=

2
50
10(356) (50)
10 1
=

50
3560 2500
9
t =

50
4.607
10.853
= =
1.7 the tabular value of t at .05 with df=9 is
1.833n (table 2) Since the computed value,
4.607 is greater than the tabular value, then
the null hypothesis is rejected.
1.8 there is a significant deference bet. The
attitudes of the employees before and after
they have seen the film.


df
.10
.05 .025 .01 .005 .0005
Level of Significance for a non-directional test
.20 .10 .05 .02 .01 .001
1 3.078 6.314 12.706 31.821 63.657 636.619
2 1.886 2.92 4.303 6.965 9.925 31.599
3 1.636 2.353 3.182 4.541 5.841 12.924
4 1.533 2.132 2.776 3.747 4.604 8.610
5 1.476 2.015 2.571 3.365 4.032 6.689
6 1.440 1.943 2.447 3.143 3.707 5.959
7 1.415 1.895 2.365 2.998 3.499 5.408
8 1.397 1.860 2.306 2.896 3.355 5.041
STUDENTS DISTRIBUTION
Level of Significance for One-Tailed test
13 1.350 1.771 2.160 2.650 3.012 4.221
14 1.345 1.761 2.145 2.624 2.977 4.140
15 1.341 1.753 2.131 2.602 2.947 4.073
16 1.337 1.746 2.120 2.583 2.921 4.015
17 1.333 1.74 2.110 2.567 2.898 3.965
18 1.330 1.734 2.101 2.552 2.878 3.992
19 1.328 1.729 2.093 2.539 2.861 3.883
20 1.325 1.725 2.086 2.528 2.845 3.850
21 1.323 1.721 2.080 2.518 2.831 3.819
22 1.321 1.717 2.074 2.508 2.819 3.792
23 1.319 1.714 2.069 2.500 2.807 3.768
24 1.318 1.711 2.064 2.492 2.797 3.745
25 1.316 1.708 2.060 2.485 2.787 3.725
26 1.315 1.706 2.056 2.479 2.779 3.707
27 1.314 1.703 2.052 2.473 2.771 3.690
28 1.313 1.701 2.048 2.467 2.763 3.674
29 1.311 1.699 2.045 2.462 2.756 3.659
30 1.310 1.697 2.042 2.457 2.750 3.646
40 1.303 1.684 2.021 2.423 2.704 3.551
60 1.296 1.671 2.000 2.390 2.660 3.460
120 1.289 1.658 1.980 2.358 2.617 3.373
infinite 1.282 1.645 1.960 2.326 2.576 3.291

Let us illustrate how to apply the
formulas for this type of
problem with the following
example:

Example: The manager of a training output for
sales representative would like to known if there might
be a significant differences among three method of
instruction (Method A, B and C) that are employed for
trainees. A different method is used for each of three
separated group of trainees. The same topic is
discussed for all groups and the same set of
examination questions are given at the end of training
seminar. A sample of examination scores of six trainees
from each of group is taken. Based on the level of
significance of 1%, test to determine if there is a
significant differences in the scores of group (that is, if
there is a significant differences in the method of
instruction employed).
Trainee
Number
Per
group
Method
A
Method
B
Method
C
1 82 80 92
2 93 86 96
3 90 78 85
4 88 82 90
5 85 76 89
6 92 84 94
Solution:
Note the differences in grade within
samples and between samples. W hat we
would like to determine, however, is
whether such differences are significant.
The first step is to solve for the
squared values of each of the grades in
the three column. Reproducing the
entries of the three X columns and get
the squared values that now form three
additional derived columns, we have:
82 80 92 6724 6400 8464
93 86 96 8649 7396 9216
90 78 85 8100 6084 7225
88 82 90 7744 6724 8100
85 76 89 7225 5776 7921
92 84 94 8464 7056 8836
2
( ) A X
A X
B
X
C
X
2
( )
B
X
530
A
X =
486
B
X =

2
( ) 94,762
C
X =

546
C
X =

2
( )
C
X

2
( ) 46906
A
X =

2
( ) 39, 436
B
X =

The next step is to get the sums of all six columns.

These are shown above. The entries in the last three
columns (the squared columns) are arrived at by
squaring the entries in the X columns, for examples:
First entries for
: 6724
:6400
:8464

Before we solve for SSt and SSc, let us identify the
following:
r = number of rows
= number of trainees per samples
= 6
2
( )
A
X
2
( )
B
X
2
( )
C
X
C = number of columns
= number of samples
= 3 (the last three columns are not include)
Degree of freedom:
dft = rc-1 dfc = c-1 dfw = c(r-1)
= 6(3)-1 = 3-1 = 3(6-1)
=18-1 = 2 =3(5)
=17 = 15
2 2 2 2
Terms for the formula of SSt:
( ) ( ) ( ) ( )
= 46,906+39,436+49,762
= 136,104
A B C
X X X X = + +


( )
=530+486+546
=1,562
A B C
X X X X = + +

2 2 2 2
2 2 2
Term for formula of SSc :
( ) ( ) ( ) ( )
(530) (486) (546)
= 280,900+236,196+298,116
=815,212
A B C
X X X X = + +
= + +


2
2
2
2
2
Substituting these values into the formulas for SSt,
SSc, and SSw we have:
( )
( )
(1562)
136,104
6(3)
= 136,104-135,546.89
= 557.11
( )
1
( )
1

6
X
SSt X
rc
X
SSc X
r rc
=
=
(

( =

=




( )
815, 212 135, 546.89
= 321.78
SSw = SSt-SSc
= 557.11-321.78
= 235.33

Substituting the computed values of SSc and SSw into
the formulas for MSSt and MSSw, we have:

SSt
MSSc
dfc
=
321.78
2
=
= 160.89

SSw
MSSw
dfwc
=
235. 33

15
=
= 15,69
Substituting the computed values of MSSc and MSSw
into the formula of Fc, we have:
MSSc
Fc
MSSw
=
160.89
15.69
=
=10.25
Let us now look for the tabular value of the F statistic (that
is, Ft) by referring to the table on F Distribution in
Appendix 4a. To identify our tabular value we need:
Two degree of freedom
A numerator value corresponding to dfc.
A denominator value corresponding to dfw.
The level of significance.
For our problem, these value are:
dfc = 2
dfw = 15
Level of significant = 1%
Base on these value, the tabular value in the table is: Ft
=6.36 (Note: These are two F table s. One for 1%,second
for 5% level of significance.)
Going through the four-step solution for test of Hypothesis, we
have:
1. 8Ho : There is no significant difference in the scores of three
groups of trainees.
2. This translates to: There is no significant differences in the
three method of instruction employed.
3. Ha : There is a significance differences in the scores of the
three group of trainees.
4. The computed value for F statistic : Fc = 10.25
5. The tabular value is : Ft =6.36
6. Conclusion
Since the computed value of F (that is, Fc =10.25) is greater
than the tabular value (that is, Ft = 6.36), we reject Ho(and
accept Ha)
Therefore: there is a significant difference in the grades of
three group of trainees
This means further that: There is a significant differences in
the three method of instruction employed.

What, then, is the implication of this
conclusion? Since there is a significant
differences in the methods, one has to
determine which method is the best and to
employ that method. Through several
applications, one is able to determine which
method will yield consistently high grades for
trainees.
Suppose the conclusion is that three is no
significant differences in the grades? Then the
three method may be considered equally
effective and any one of the method may be
used for training purposes.
One Way Classification for
Unbalanced , Array
In an unbalanced array, the samples do not have the
same array of observations the formulas are slightly
different from those used in a balanced array. Let us
summarized for an unbalanced array. These are:



Note: the formulas of the other term remain the same.
These are:

( )
2
2
( )
X
SSt X
n
(
( =
(

(



2
2
( )
( )
X
X
SSc
r n
(
(

=
(

.
1 dft n =
1 dfc c =
dfw n c =
SSc
MSSc
dfc
=
SSw
MSSw
dfw
=
MSSc
Fc
MSSw
=
The slight differences compared to the
formulas of a balanced array are:

The denominator for the second term of SSt
and SSc is n instead of rc.
In the first term SSc, r appears inside the
brackets instead of outside. This means r is
distributed into each of the entries.
dft is n-1 instead of r c-1
dft is n-c instead of c(r-1)
2
( ) X

Note: the formulas of the other terms remain

the same. These are:
SSw MSSc Fc
dfc MSSw

The values of all the terms will again be solved
by following the same sequence as in the case of a
Balanced array.
Let us now apply the formulas for an
Unbalanced array in the following example. The
example is essentially the same as in the preceding
example of a Balanced array except that we
introduce changes in the number of observations
for examples B and C, that is:
Sample A : the same 6 observations
Sample B : the first 5 observations only
Sample C : the first 4 observations only
This means that only 5 and 4 trainees
respectively are taken for samples B and C
Thus the column entries are:

82 80 92 6724 6400 8864
93 86 96 8649 7396 9216
90 78 85 8100 6084 7225
88 82 90 7744 6724 8100
85 76 __ 7225 5776 ____
92 __ __ 8464 ____ ____
__ __ __ ____ ____ ____
530 402 363 46,906 32,380 33,005


A X B X C X
2
( ) A X
2
( ) B X
2
( ) C X
Identifying the values of terms
in the formulas, we have:
n = the number of observations in the three
samples
= observations in samples A, B, and C together
= 6+5+4
=15
Degrees of Freedom

dft = n-1 dfc = c-1 dfw = n-c
= 15-1 = 3-1 = 15-3
= 14 = 2 = 12

Terms in the formula SSt
=
= 46,909 + 32,380 + 33,005
= 112,297

=
= 530 + 402 + 363
= 1295


2
( ) X

2 2 2
( ) ( ) ( ) A B C X X X + +

( ) X

A B C X X X + +

Terms in the formula SSc
.

( )
2
( )
2
The first term in the formula of SSc is
This means that we should first determine the value
for each sample after which we get the sum
of these computed results for the three samples
X
r
X
r

( )
2
2
(530) 280, 900
Sample A = 46,816.67
6 6
A X
r
= + +

( )
2
2
(402) 161, 604
Sample B: = 32,320.80
5 5
B X
r
+ + =

( )
2
2
(363) 131, 769
Sample C: = 32,942.25
4 4
C X
r
+ + =

Therefore, the first term of the formula for SSc is:

Substituting these computed values into the
formulas of SSt and SSc, we have:
( ) ( ) ( ) ( )
2 2 2 2
6 5 4
A B C X X X X
r
(
(
= + +
(

( ) ( ) ( )
2 2 2
530 402 363
6 5 4
= + +
46,816.67 32,320.80 32,942.25 = + +
112, 079.72 =
( )
( )
2
2
t X
X
SS
n
=
(

1, 677, 025
112, 297
15
=
112, 297 111,801.67 =
495.33 =
( )
( )
2
2
X
X
c
n
r
SS
(
(
(
(


=

112, 079.72 111,801.66

278.06
=
=
Solving now for SSw, MSSc, MSSw, and Fc, we have:
W SS SSt SSc =
495.33 278.06 =
217.27 =
278.06
139.03
2
SSc
MSSc
dfc
= = =
217.27
18.10
12
SSw
MSSw
dfw
= = =
139.03
7.68
18.10
MSSc
Fc
MSSw
= = =
Let us now identify the tabular value, Ft, based on the
following information:
Degree of Freedom numerator = 2
Degree of Freedom, denominator = 12
Level of Significance = 1%
From the table of F Distribution in Appendix
4a, Ft=6.93. Going through the four-step
solution, we have :
1. Ho: There is no significant difference in the grades
of the three groups of trainees.
This means that: there is no significant difference in
the three methods of instruction.
Ha: There is significant difference in the grades of
the three groups of trainees.
3. The tabular value is : Ft = 6.93
4. Conclusion:
Since the computed value (that is, Fc = 7.68)
is greater than the tabular value (that is, Ft =
6.93), we reject Ho (and accepts Ha)
Therefore: There is a significant difference in
the grades of the three groups of trainees.
Therefore: There is a significant difference in
the three methods of instruction employed.

2. The computed value is : Fc= 7.68
Table 3. CRITICAL VALUES OF F
Fractions for .05 (lightface) and .01 (boldface)
Two-Factor Analysis of Variance
In the contingency Table 15.1 a schematic representation of a two-
factor ANOVA is presented. In this table we have a design with
treatment groups of size 10(N) and 2 factors: a column factor, C
and a row factor R the column factor has 4 levels and the row
factor 3. In the table 15.1 there are four columns, c=4, and three
rows, r=3. This causes the table to contain 12 cells. Each of these
cells correspond to a treatment group and each contains a sample
that has been uniquely treated the ten individuals in the upper-
left hand cell received the treatment associated with the level 1 of
factor R and level 1 of factor C.

The effects of the treatment combination r and c will be shown
when the size of the mean of this cell is compared with the means
of the 11 other treatments. The effects of an entire factor, called
factor main effects, are calculated through a study of the marginal
means. That is, the differences between the marginal means



1, 2, 3, 4 C C C C X X and X X
Table 15.1 MODEL OF TWO-FACTOR
ANOVA DESIGN
Factor C: Column
LevelC1 Level C2 Level C3 Level C4

Level 1 r Nr=40
Xr
Level 2 r Nr=40
Xr
Level 3 r3 Nr=40
Xr

Factor R: Nt=120
Rows Xt
N11 =10
X11
N12=10
X12
N13=10
X13
N14=10
X14
N21=10
X21
N22=10
X22
N23=10
X23
N24=10
X24
N31=10
X31
N32=10
X32
N33=10
X33
N34=10
X34
Show the main effects of fator C across all levels of
factor C. Similarly, differences between the marginal
row means when analyzed, show the
main effects of factor R.
Partitioning the Sum-of-Squares
The usual formula for the total sum-of-squares is
again used here:

The within sum-of-squares is obtained by the
following formula:


Where is the number of cell groups and N is the
number in each cell.
1, 2, 3, r r r X X X
k
( )
2
2
t
t
t t
N
SS
EX
= EX
2
2
k
w t
N
SS
| |
E EX
|
\ .
= EX
SS is not computed directly in this type of analysis;
instead, it is partitioned into the following: the
main effect of C(SS), the main effect of R (SSr)
and the interaction effect (SS). Thus:


The computational formulas for these follow:




Where and are the sums of the X values
in each column and row respectively, and N and
N are the sample sizes of each column and row.


c EX
r EX
. b c r c r SS SS SS SS = + +
( ) ( )
2 2
c t
c t
c
N N
SS
E EX E EX
=
( ) ( )
2 2
t r
r t
t
N N
SS
E EX E EX
=
The SSc r is found by subtraction:


After all the sums of squares have been
determined, the usual ANOVA table is set up and
each SS is divided by its appropriate degrees of
freedom to obtain the corresponding mean
square (MS) or variance. The test of the main
effects of factors C and R and the interaction
effect are then evaluated by making an F test for
each by dividing each MS by the MSw. This latter
mean square is sometimes referred to as the error
term.
c r t w rc r SS SS SS SS SS =
Degrees of freedom for this type of problem are determined
as follows:


Where k is the number of samples and N is the size of each
sample, assuming Ns are equal.
( ) ( )
1
1
1 1
col c
row r
c r c r
df
df
df
=
=
=
( ) 1
1 t t
w k
df N
df
=
= N
: b df
TABLE 15.2 Scores of 80 individuals of 80
th
sexes on an
attitude scale after the administration of four different
films
FILM
1 2 3 4
MALES
10 14 13 16
8 12 12 12
.6 8 10 10
4 4 9 9
4 4 9 9
4 4 7 7
2 3 4 7
2 2 4 6
2 2 4 5
1 1 2 5
=43 54 74 86 =257







FEMALES
14 14 10 18
12 13 10 16
12 12 10 15
10 11 9 14
8 10 9 13
6 9 7 10
4 7 7 10
2 6 6 10
2 3 5 10
1 2 4 9
=71 87 77 125 =360
c=114 141 151 211 c=617
An Example
In table 15.2 are the responses of 40 female and 40 male
university students to a attitudinal scale. Each group of 40
was randomly divided into two groups of 10 each then
each of these groups of 10 was shown one of four
different films on a controversial subject. Later the
attitudinal scale on this subject was administered to each
individual. For the study, alpha was set at .05.
First the scores in each of the columns are summed, as
shown in the table, then the two rows are summed. Both
of these sums come to 617. The is obtained by
squaring each of the separate scores and summing the
squares. The squares and sum can most readily be
obtained by the student using a pocket calculator
The resulting is 6159. Then
we proceed to obtain the various sums-of-squares:

2
EX
2
EX ( )
2 2 2 2
10 8 6 ....9 + + +

(1)
( )
2
2
t
t t
t
SS
N
=
EX
EX
( )
2
617
6159
80
=
380, 689
6159
80
=
6159 4758.6 =
1400.4 =



(3)
( )
2
2
w t
N
SS
k E EX
= EX
2 2 2 2 2 2 2 2
43 54 74 86 71 87 77 125
10
6159
+ + + + + + +
=
51,801
6159
10
=
6159 5180.1 =
978.9 =
( ) ( )
2 2
c t
c t
c
N N
SS
E EX E EX
=
( )
2
2 2 2 2
617
114 141 151 211
20 80
+ + +
=
100,199 380, 689
20 80
=
5009.95 4758.61 =
251.34 =
(2)




(5)
( ) ( )
2 2
t r
r t
r
N N
SS
E EX E EX
=
( )
2
2 2
617
257 360
40 80
+
=
195, 649 380, 689
40 80
=
4891.22 4758.61 =
132.61 =
c r t w c r SS SS SS SS SS =
1400.4 978.9 251.34 132.61 =
37.6 =
(4)
The number of degrees of freedom for the various
parts of this problem are:
Then the data entered into ANOVA table:
1 80 1 79 t t df N = = =
( 1) 8(10 1) 72 w k df N = = =
1 4 1 3 c c df = = =
1 2 1 1 r r df = = =
( ) ( ) ( ) ( ) ( ) 1 1 4 1 2 1 3 1 3 c r c r df = = = =
Source of Variation SSc df MS F P
Between
Columns 251.34 3 83.78 6.16 <.05
Rows 132.61 1 132.61 9.75 <.05
Interaction 37.6 3 12.53 .92 >.05
Within 978.9 72 13.6
Total 1400.4 79
The following decision model has been built for
these data
Not

Not
Interaction effects are not present.
Interaction effects are presents.
0 1 2 3 : c c c H = =
1 : H
0 H
0 1 2 3 : r r r H = =
1 : H 0 H
0 : H
1 : H
: .01 o
c : 0 DR Reject if 5.49;otherwise, do not reject. obs H F >
r : 0 DR Reject if 5.49;otherwise, do not reject. obs H F >
c r : 0 DR Reject if 4.11;otherwise, do not reject. obs H F >
In this problem for columns, teachers, exceeds the .01
table value of for This leads to the
rejection of and to the conclusion that the main
effect factor, teacher, is significant. On the other hand,
for texts is less than the 0.01 table value of
So in this case is accepted and we conclude that the
interaction effect is also significant at the .01 level. The
presence of significant has to be acknowledged when
taking about the significant main effect for teacher.
From the data in Table 15.3, it appears that students of
teacher 2 using text 1 did better with those with the
other combination of factors. Methods exist for carrying
out these post hoc comparisons (such as a test for
simple effects and individual comparisons); however,
these are too complicated for an elementary text.
obs F
5.49 F =
of 2.27. df
0 H
obs F
for of 2.27. F df
0 H
Another Application of the Two-Factor
Design
In some cases there are two treatments or experimental variations
and only one observation of each combination of factors. We
shall illustrate this type of design using the data in Table 15.4
where the ratings of 6 trainees by 4 supervisors are presented.
The methodology presented below is similar so that the given
before, except that the sum of squares related to interaction
and to error effect are not separated or partitioned.
Note that in Table 15.4 various squaring and summing have
carried out. This illustration is one of the simpler two-way
ANOVA applications in that although there are two variables,
rater and rate, there is only one observation recorded in each
cell or for each combination of the two variables. Such a model
is referred to as a random model, since both trainees and
supervisors are assumed to be random samples from normal
populations.
Table 15.4
The ratings of Six Trainees By Four Supervisors
Rater Squares of Ratings
Trainee A B C D A B C D

1 10 6 8 7 31 100 36 64 49
2 4 5 3 4 16 16 25 9 16
3 8 4 7 4 23 64 16 49 16
4 3 4 2 2 11 9 16 4 4
5 6 6 6 7 27 36 64 36 49
6 9 7 8 7 31 81 49 64 49

40 34 34 31 139 306 206 226 183
=921
of
Rows
E
.
The analysis of variance proceeds as follows:
1) The total sum-of-squares (SSt) is found in the
usual manner. ( )
2
2
2
139
921
24
921 805
116
t SS =
EX
EX
N
=
=
=
(2) The sum of the squares for rows, trainees(SS?) is
next obtained.
( )
2
2 2 2 2 2 2
139
31 16 23 11 27 31
4 24
3557
805
4
889.25 805
84.25
r SS
+ + + + +
=
=
=
=
(3) The sum of the squares for columns, Raters (SSc),
is then calculated.
( )
2
2 2 2 2
139
40 34 34 31
6 24
c SS
+ + +
=
4873
805
6
=
812.16 805
7.16
=
=
(4) After the SS for both rows and columns have been
obtained. This leaves the residual or error sum squares, SSc.
The error term is a combination of interaction and error
effects.
( )
( )
116 84.25 7.14
116 91.41
24.59
e t r c SS SS SS SS = +
= +
=
=
(5) Finally a table set up:
Source of Variation df SS MS F P
Rows(trainees) 5 84.25 16.85 10.27 <0.01
Columns(raters)3 3 7.16 16.85 1.46 >0.01
Error 15 24.59 1.64
Total 23 116.00
The first F in this table is obtained by dividing the
variance for rows by the error variance
This obtained F of 10.27 is highly significant for
5.15 degrees of freedom, and is far beyond the 1
percent point. This indicates that the supervisors
did discriminate among trainees. The second F is
obtained by dividing the variance for columns,
Raters, by the error variance . With 3.15
degrees of freedom the obtained F of 1.46 is not
significant, indicating that the supervisors did not
differ significantly among themselves in their
ratings of the trainees. In other words, their
ratings were reliable.
( )
/ r e MS MS
( )
/ c e MS MS
Data from this technique are often used in
determining the reliability of raters. Reliability is
discussed in some detail in Chapter 18. By the use of
the following formula, the reliability of the ratings of
a single rater is obtained:


where k is the number of columns and the other terms
are as used above:
( )
1
r e
r e
MS MS
r
MS k MS
EX

=
+
( )
16.85 1.64
16.85 4 1 1.64
r
EX

=
+
15.2
16.85 4.92
15.21
21.77
.70
=
+
=
=
The reliability of all four raters taken together is
obtained by:
r e
tt
r
MS MS
r
MS

=
16.85 1.64
16.85

=
15.21
16.85
=
.90 =
In this chapter we have described only a few
of the simple uses of the analysis of variance.
To go further is beyond the scope of this
book. The interested student will find in more
advanced texts the large number of
experimental designs that can be analyzed by
the application of the analysis of variance.
In this chapter we have described only a few of the
simple uses of the analysis of variance. To go further
is beyond the scope of this book. The interested
student will find in more advanced texts the large
number of experimental designs that can be
analyzed by the application of the analysis of
variance.
Prepared by;
Wily A. Aglobo
John Leonard Calinog
Fevilyn Dimasuay
Karen Joy D. Elchico
Mary Grace Maxilum