[R] Transformed predictor in linear model

[Francisco Vergara]

Hi

I fitted a mixed model with several categorical and continuous fixed 
variables and 3 nested random intercepts, using a transformation of the 
predictor y = log(y+1) and I used the defaults contrast treatment and 
contrast polynomial.

The resulting coefficients are in log units meaning that the difference 
between the categorical (Factor) variables builted by Contr.treatment is 
expressed on log values but to retransform them to actual (y) values is not 
as straight forward as to exp(beta)-1.  I am sure that this is a problem 
that you guys face every day but I can't find an easy way around this. The 
easiest way to go is to report the betas in transformed values but this is 
not very useful for the practical purposes of the study.

Any comments??

_________________________________________________________________
Want to check if your PC is virus-infected?  Get a FREE computer virus scan 
online from McAfee.