I've collected data from school children in year 7 (11-12yrs) and year 9 (13-14yrs). I want to see if there is a significant difference in physical activity between 'teacher rated ability groups' (set 1, high ability and set 4 low ability).

I have a score for physical activity for each student which ranges from 0 (no activity) to 50 (really active). The histograms I have produced are an L shape with most scoring 0 and only 1 or 2 scoring 40-50.

I also have a number of outliers, especially in the year 9 data where very few are active. I don't want to delete cases as they are correct and from the intended population.

I have tried log transforming the data and this worked with the year 7 data but not the year 9. Can anybody offer some advice as to what I should do?!

My groups are also unequal size, would that make a difference?
Last edited:


TS Contributor
How unequal are the sample sizes?

You might consider a nonparametric test, such as Mann-Whitney (the nonparametric analogue to the Student's t-Test).