Power transformation

Home > Statistics Every Writer Should Know > The Stats Board > Discusssion

Power transformation
Message posted by Milind Sathye (via 203.22.86.226) on March 1, 2001 at 1:04 AM (ET)

I am dealing with two dependent variables that are bounded by 0 and 1. For Box M (under MANOVA) to be insignificant, there is a need to transform these variables. I have tried all sorts of transformations each resulting gives significant Box M. Only when the DVs are transformed to the power of 60 and above, does the Box M gives insignificant results. Since transformations are generally done to the pwer of say 2 or 3 at the most, my worry is whether transformation of DVs that I have done is valid statistically? If not what is the alternative?

READERS RESPOND:
(In chronological order. Most recent at the bottom.)

Re: Power transformation
Message posted by Milind Sathye (via 203.22.86.226) on April 28, 2001 at 6:43 AM (ET)

I have found out that power transformation can be done at the most to the power of 2 or 3. In cases as I have refereed the appropriate transformation could be logit. If nothing works then to get a non significant Box M one should follow the procedure as given in the book by Tabachnick and Fidell on Using Multivariate Statistics. The book says that where the groups are of approximately equalsize, even if Box M is significant, it may be ignored. The results continue to be robust.
I have added this posting to help others who may face similar problem in their work.
Thanks

Your $5 contribution helps cover part the $500 annual cost of keeping this site online.