FW: Will's questions/comments on GLM & transformations
- I inadvertently sent my reply to Dwight only and not the whole group.
I hope my attempt to be helpful with the Tabachnick and Fidell reference succeeds.
Some useful suggestions and in support, Tabachnick and Fidell in their 1996 Using Multivariate Statistics, 3rd edn recommend
transformations for different types of non-normally distributed data. Page 83 has a particularly useful figure.
In short, as you rightly indicate, log-transformations are not necessarily the ones to use.
Professor Edward M Winter
The Centre for Sport and Exercise Science
Sheffield Hallam University
Collegiate Crescent Campus
SHEFFIELD S10 2BP
Tel 0114 225 4333 (International +44 114 225 4333)
Fax 0114 225 4341 (International +44 114 225 4341)