Course Changes and Future Plans
The study reported here took place in the spring semester, 2001, and I have taught E538 once since, in spring 2002. In response to what I learned from this study, and to some extent for other reasons, I made a few changes between those two semesters:
· Because of the responses on Question 2 of the posttest, I explained paired t tests more carefully in 2002. Judging from exam results that year, I think I can improve my explanations of this topic even further.
· The posttest results on Question 3 were also not fully satisfactory, so in 2002 I emphasized the commonness of non-normality, effectively, I believe.
· I discussed the background for Question 4 a little more in 2002 than in 2001, but I think it would be useful to discuss the issue even more in the future.
· Questions
5, 6, and 12 each addressed correct interpretation of
statistical significance tests, in various ways. Although I was generally successful in teaching that material in
2001 (least so with Question 6), I continued to stress this material in 2002,
and added homework assignments to improve student understanding of this
important issue. I will continue to
refine my teaching of that issue in the future.
Between the two years, I published a paper on these issues (Parkhurst, 2001), and in 2002 assigned that paper as
required reading to provide reinforcement and additional examples of what I
discussed in lecture.
· I was not satisfied with the students’ understanding of statistical power, based on Question 9 of the posttest in 2001. I tried to teach this more effectively in 2002, but ran out of time near the end of the semester, and did not do as well as I would have liked. I will try to teach this subject better in the future.
· Question 10 dealt with the advantages of Bayesian analysis over conventional frequentist statistics, and showed that I had been fairly successful in dealing with that topic in 2001. In 2002, we used Matlab rather than SPSS for statistical computations; that allowed easier calculations in Bayesian analysis, so I was able to assign more complete problems of this nature in 2002.
· In response to the odd results with Question 13, I made a point in 2002 of explaining that resampling methods cannot add data to an inadequate dataset; rather, their advantage is to free the analyst from having to assume a specific distribution form (such as normality) for the data under consideration.
· Unrelated to the 2001 pretest-posttest results, I had the students use the “statistics toolbox” in Matlab rather than SPSS, for statistical computations. This made simulations much easier, and I think students learn a lot from these. On the other hand, some students complained on the 2002 semester-end course evaluations about not learning a “standard” statistical package like SPSS or SAS. I may revert to SPSS in the future, though I haven’t made a final decision about this. (For 2003, I plan to use “R,” a free clone of the “S” statistical language, and similar to the commercial package, S-plus.)
· In
2001, 10% of the course grade was based on homework assignments (involving
calculations done by hand or with SPSS software), which students were supposed
to complete alone. Near the end of the
semester, it became clear that at least a few students were working together on
this material. In 2002. I had the
students do homework in groups of three, with each group having different
numbers to prevent using, or comparing with, another group’s solutions. My main purposes for this change were to
give students experience in working together, and so they could learn from each
other as they worked. The students
seemed to like this approach, and I thought it worked well, so I plan to use it
in the future.
Of the five 2002 assignments, four involved using the Matlab statistics toolbox
to perform calculations and required subsequent discussion and interpretation
of results. The middle one was
different¾it
asked the students to read documents from two environmental regulatory
agencies, and to comment on whether the statistical methods outlined in those
two contrasting documents protected human health or not.
©2003, D.F. Parkhurst