How to resolve factor-specific variance of random effect in lme

I assume that the variances of random effects in my mixed effect model will be different for different levels of a fixed coefficient BTyp.

Here is my model

fm2 <- lme(CA ~ 1 + pF+Tiefe+BTyp+Tiefe:pF+BTyp:pF, data=data2, 
           random = list(~ 1 + pF|Probe))
fm2_Btyphet<-update(fm2, weights=varIdent(form=~1|BTyp))

I managed to include BTyp-specific variances for random effects using the function lmer, but this function does not allow us to consider the heterogeneity of variance within the group error (which is better to consider in my case). My question is, how do I enable “Btyp” special variances for random effects using a function lme?

Below you can see how this works with the function lmer.

CA ~ 1 + pF + Tiefe + BTyp + Tiefe:pF + BTyp:pF + 
     (0 + Pind + pF | Probe) + (0 + Bind + pF | Probe) + (0 + Tind + pF | Probe) 


 Data: data2 

 AIC   BIC logLik deviance REMLdev

   21987 22092 -10975    21979   21951

Random effects:
 Groups   Name Variance Std.Dev. Corr 

 Probe    Pind 158.6058 12.5939         
          pF     2.4289  1.5585  -1.000 

 Probe    Bind 134.6383 11.6034         
          pF     2.7619  1.6619  -1.000 

 Probe    Tind 490.6714 22.1511         
          pF    46.3533  6.8083  -1.000 

 Residual      316.9860 17.8041    

Number of obs: 2530, groups: Probe, 45

Pind, Bind, Tind BTyp.

+4
1
+1

Source: https://habr.com/ru/post/1524093/


All Articles