Information Technology Reference
In-Depth Information
Table 10.1.
Raw data for a (five-group) one-way within-subjects ANOVA
Factor
A
: Symptom intensity measured before and after drug therapy
One month One week
One week
One month One year
before
before
after
after
after
Subjects
a
1
a
2
a
3
a
4
a
5
Sum (
S
)
s
1
12
12
7
5
5
41
s
2
9
10
5
6
5
35
s
3
9
8
7
5
6
35
s
4
8
6
6
4
4
28
s
5
8
9
6
4
5
32
s
6
9
10
8
6
7
40
s
7
12
10
7
5
4
38
s
8
6
7
5
7
5
30
M
=
9
.
125
M
=
9
.
000
M
=
6
.
375
M
=
5
.
250
M
=
5
.
125
in other research approaches, such as archival research where the data are
alreadycollected,thebestwemayoftenbeabletodoistocomparedifferent
cases whose dependent variable scores are recorded under different levels
of the independent variable.
10.6 A NUMERICAL EXAMPLE OF A ONE-WAYWITHIN-
SUBJECTS DESIGN
The simplified data set that will serve as our example uses a time-based
independent variable in a pretest-posttest configuration. In our hypo-
thetical study, it is hoped that a certain drug will alleviate the intensity
of symptoms of a certain disease. Symptom intensity was evaluated on
a twelve-point scale with values toward twelve indicating more intense
symptoms. Two pretreatment baseline measurements are made, the first
a month prior to treatment and the second a week prior to treatment.
Posttreatment measures of symptom intensity are made after one week,
one month, and one year following administration of the drug. In this
simplified example, we have eight patients in the study. As can be seen
from the means of the conditions in Table 10.1, symptom intensity hovers
around a value of 9 in the pretests, drops to a value somewhat in excess of
6 after the first week, and then appears to level off at a value just a little
more than 5 after a month.
10.7 EFFECT OF INTEREST IN THIS DESIGN
10.7.1 OMNIBUS EFFECTS
In a within-subjects design, the total variance of the dependent variable
is partitioned into between-subjects and within-subjects variance. The
between-subjects variance indexes individual subject differences, and is
ordinarily not of interest other than having it account for its portion of
the variance.
