[R] Mixed model specification

Boks, M.P.M. M.P.M.Boks at umcutrecht.nl
Thu Jun 11 02:57:32 CEST 2009


Dear R-friends,
 
I would very much appreciate any direction you can give me regarding the following:
 
I am trying to specify a model with three random effects (var1, var2 and constant) in two levels (nested in pair):
 
 
lme(dependend ~ age + gender, random=list(pair = pdIdent(~var1 + var2 +1)), data=data, method = "ML", na.action="na.omit")
 
this works, the equivalent with formula also works (although only for method=" REML"??)
 
lme(dependend ~ age + gender, random=~var1 + var2 + 1|pair, data=data, method = "REML", na.action="na.omit")
 
Now my problem: how do I specify a random effect with var1 nested in individuals and var2 and constant nested in pair:
 
lme(dependend ~ age + gender, random=~var1 + (var2 + 1|pair), data=data, method = "REML", na.action="na.omit")
 
Is not working (invalid formula for groups).
 
I am trying to specify a mixed effect model as outlined in biometrical Modeling of Twin an Family Data Using standard mixed model software, Rabe-Hesketh et al , Biometrics 2008, 64:280-288
 
Any suggestion would be much appreciated.
 
Marco 




More information about the R-help mailing list