The groups are small and we have interval data. We could try a transformation, such as the log, and then use the two-sample t method. We could also use the Mann-Whitney U test (Intro 14.3). After the problem with their analysis had been pointed out, the authors used the latter approach. This, not surprisingly, was highly significant, P=0.0002 (Hales and colleagues, unpublished personal communication to Lancet).
Back to question.
This page maintained by Martin Bland.
Last updated: 1 July, 2005.