Internet Learning Volume 3, Number 2, Fall 2014 | Page 63

Internet Learning
Table 2. Regression models predicting final course grades as a function of gender
and self-efficacy
Model
1
Model
2
Gender 4.69* 2.45
Self-efficacy
(pretest)
3.82**
R 2 0.07 0.15
RMSE 8.49 8.19
* p < 0.05, ** p < 0.01, *** p < 0.001
final course performance with some reasonable
level of accuracy. This result suggests
that instructors ought to make this information
known to students, ideally through
a computerized early detection system that
automatically alerts instructors when the
model predicts that a student may fail the
course. Such a system ought to also be available
to students, both to help students stay
on track in the course and to help students
learn and internalize the important metacognitive
skills of self-monitoring.
Our analysis of the relationship between
gender, self-efficacy, and course performance
suggests a different understanding
of the gender gap in physics. Female students’
lower levels of self-efficacy (a mean of
3.2 on a 5-point scale, compared to 3.8 for
their male counterparts) suggest that self-efficacy
differences may be at least partly responsible
for the gender gap. This suggests
that an important next study is to examine
in detail what factors lead to the gender gap
of self-efficacy in science.
The success of the very short (seven-item)
self-efficacy measure suggests that
there may be other noncognitive characteristics
that might also be predictive of later
student performance. We were surprised to
discover that students’ self-efficacy of their
performance in the Peer Instruction environment
did not help to predict students’
final grades, especially since both general
self-efficacy and students’ actual performance
on the ConcepTests both were highly
predictive. One avenue of future work is to
refine our instrument for measuring Peer
Instruction self-efficacy so that it might be
more predictive of final grades. Another
is to examine other noncognitive abilities
that can be measured early on and that are
predictive of course outcomes (e.g., study
skills and habits, attitudes towards learning
and the discipline, etc). Even though
our analysis was retrospective and does not
demonstrate causality between self-efficacy
and course outcomes, the results do suggest
that the development of an intervention to
help improve students’ self-efficacy may be
worthwhile, especially for women. Further,
given that our study was conducted in one
classroom at one institution, future work
that replicates and expands on these findings
across a range of disciplines and institutions
would be valuable in helping to shape what
a successful intervention would look like.
62