[R-sig-ME] Skewed residuals from random effects

[Eric Edeline]

Dear list,

what to do when a normally-distributed random effect is justified by the 
data structure, drops model AIC but badly skews the residuals? And 
actually how can this happen? I am modeling body length of 50 fish 
species as a function of various environmental variables (see below). 
Any feed back would be greatly appreciated.

Thanks!


My model:
m046<-lme(log(Length) ~ 
log(Slope)+log(Width)+log(D)+log(Compint2)+log(Pred2)*log(Temp)+Year, 
data=Data, na.action=na.omit,
random=list(Strategy=pdDiag(form=~1), Method=pdDiag(form=~1), 
Species=pdDiag(form=~1)))    #normal random effects but skewed residuals
                                                                         
                                                                         
                        #residuals from a lm are nicely distributed


-- 
Eric Edeline
Assistant Professor
UMR 7618 BIOEMCO
Ecole Normale Supérieure
46 rue d'Ulm
75230 Paris cedex 05
France

Tel: +33 (0)1 44 32 38 84
Fax: +33 (0)1 44 32 38 85

http://www.biologie.ens.fr/bioemco/biodiversite/edeline.html