Michael,
It is very unlikely that any transform will give you satisfactory
results with a doubly bounded response. Instead of transforming the
data, you should consider taking the bounded nature of the response
scale into account from the beginning, i.e. choose a distribution model
suited to the analysis of bounded response.
If you are in the context of regression, then use beta regression:
http://cran.r-project.org/web/packages/betareg/index.html
If you are more in an item response theory context, then use Beta
Response Models:
http://apm.sagepub.com/content/31/1/47.short
Best regards,
Yvonnick Noel
University of Brittany, Rennes 2
France
---
You are currently subscribed to tips as: [email protected].
To unsubscribe click here:
http://fsulist.frostburg.edu/u?id=13090.68da6e6e5325aa33287ff385b70df5d5&n=T&l=tips&o=30143
or send a blank email to
leave-30143-13090.68da6e6e5325aa33287ff385b70df...@fsulist.frostburg.edu
