Of 18 students originally, I had 16 finish the course. Excluding the 2 who dropped from the data, the normalized gain on the FCI using class averages of the tests was 0.57
I had 1 student who made a 30 on the pre-test, but missed 1 on the post-test (and he knew which question when I told him), but the normalized gain would have a 0 in the denom. Excluding him, the average of the individual gains was 0.62 with s=0.19.
The smallest gain was 0.32 from my worst student (went from 5 correct to 13 correct). The largest was 0.96 ( 7 correct to 29 correct), but he's still poor at actually solving problems.
The average score went from 14.0 (s=7.4) to 23.2 (s=5.3) excluding the 2 drops. Of the 16, 8 scored <20 on the pre-test and only 2 were <20 on the post.
I taught the course using a SCALE-UP model with lots of peer-interaction and in-class experiments.
I think I can safely say that there was an improvement in concepts. For the better students, the improvement was more dramatic.
Unfortunately, I only have 1 FCI data set data from when I taught the class traditionally. In that I had 13 students who finished. The gain based on average scores was 0.65, and the average of the individual gains was 0.65 (s=.16). 10 students <20 on the pre-test and only 1<20 on the post.
While there's not a big difference in these two data sets, the students in the interactive setting were much more alert and enjoyed the class more.
I should add that most of these students were engineering majors.
I didn't have time to administer the Lawson test (Scientific Knowledge)