Statistical methods and the evaluation of school-based AIDS education in Africa (reply)

Abstract

We are grateful for the opportunity to re-analyze our data as suggested by Kinsman et al. and we apologize if the P values quoted in our paper appear to claim unwarranted precision. A paired t-test approach weighting the mean % scores proportionately to the differing numbers of students involved gave the following results. As expected this analysis gives results that have a higher P value than in the original analysis. The differences between changes in knowledge scores become of borderline significance but between changes in attitude scores are still highly significant. The behavior score changes still do not differ by more than could easily be explained by chance. So the conclusions in their paper remain substantially unchanged but may now be assigned a higher validity. Bearing in mind the limitations of our study we hope that these results will provide encouragement to those responsible for AIDS education programs in Africa. (full text)