COMPUTER ANXIETY AND COMPUTER SELF-EFFICACY
Inhaltsverzeichnis:
- Title
- Abstract
- Introduction
- Method
- Results
- Discussion
- References
- Appendix A
- Appendix B
- Table 1
- Table 2
- Footnotes
- Contributor
Results
The item sets for CAX and CSE were analysed test-
theoretically on the basis of the pretest data and with
respect to the Partial Credit Model (PCM) of Masters (1982).
The PCM was developed by generalizing the Rasch Model to
items with more than two response categories. In contrast to
the model of Classical Test Theory, which is usually applied
in this context, the PCM has three important advantages: 1)
it explicitly refers to a limited range of response
categories for each item; 2) it allows empirical testing of
whether the reponse categories are actually ordered
ordinally with respect to the variable to be measured; 3) if
it fits to data then it provides a solid empirical and
theoretical basis for measurements - at least - on interval
scale level (cf. Rost, 1996). All test-theoretical analyses
were performed with WINMIRA (see Davier, 1997).
The two item sets for CAX and CSE were analysed
separately. Items with disordinal response categories or
with a bad fit to the PCM were eliminated stepwise. For CAX
13 items had to be eliminated. The remaining 7 items are
'Thinking about taking a computer course', 'Applying for a
job requiring computer training', 'Sitting in front of a
home computer', 'Learning to write programs', 'Learning
computer terminology', 'Reading a computer manual', and
'Taking a class in the use of computers'. The reliability
for these seven items is 0.84. To analyse the CSE-items, the
two easier response categories had to be comprised into one,
because for one item nobody chose the easiest category.
Thus, the PCM reduces to the ordinary Rasch Model. All nine
items correspond extremely well to this model. No item
needed to be eliminated. The reliability for the nine items
is 0.83. On the basis of the remaining items, pretest scale
values for CAX and CSE were determined by weighted maximum
likelihood estimation (Warm, 1989). The correlation between
both scales is -0.50 (p<0.001; two-tailed) and -0.60 after
correction for attenuation; i.e. CAX decreases strongly with
CSE. Moreover, nearly all further questionnaire variables
indicating previous computer use or computer performance
correlate negatively with CAX and positively with CSE.
Subjects participating in both tests have no
significantly different pretest-values in CAX or CSE than
subjects participating only in the first test. Nor is there
any significant interaction between posttest participation,
computer lecture and/or computer course attendance. This
indicates that neither CAX nor CSE influenced course
dropout. For the 41 pretest-posttest subjects, neither
computer lecture attendance, nor computer course attendance,
nor the interaction between both produced a significant
effect on pretest values of CAX or CSE. This indicates that
the four groups in the factorial design are at least roughly
comparable. With scale values normed to zero as the lower
and one hundred as the upper scale limit, the total pretest
means of the pretest-posttest subjects are 38.01 (sd=14.85)
for CAX and 48.49 (sd=16.15) for CSE. As both values lie in
the middle of the scale, possible changes can be detected.
The posttest values of the 41 pretest-posttest subjects
were determined by assigning measurement values to raw
scores according to the assignment function that was
calculated according to the PCM for the 249 pretest
subjects. Recourse to pretest data was necessary because the
posttest data were too few to allow a reasonable test-
theoretical analysis in its own right. The changes in CAX
and CSE were determined by subtracting pretest from posttest
values (see table 2). CAX reduced both for students who
attended the computer course and for those who did not;
however, it reduced significantly more for students who did
(F(1,37)=6.185, p<0.05). CSE enlarged for students who
attended the course and reduced slightly for students who
did not. The difference between the changes is significant
(F(1,37)=4.915, p<0.05). Neither lecture attendance nor the
interaction between course and lecture attendance produced a
significant effect; - neither for CAX, nor for CSE. There
was even a non-significant increase of CAX for subjects who
attended the lecture but did not attend the course.
Preview:Method
Next:Discussion