# [R] fixed and random effects in lme

Federico Calboli f.calboli at ucl.ac.uk
Thu Feb 13 20:57:03 CET 2003

```Hi All,

I would like to ask a question on fixed and random effecti in lme. I am
fiddlying around Mick Crawley dataset "rats" :

http://www.bio.ic.ac.uk/research/mjcraw/statcomp/data/

The advantage is that most work is already done in Crawley's book (page 361
onwards) so I can check what I am doing.

I am tryg to reproduce the nested analysis on page 368:

model<-aov(Glycogen~Treatment/Rat/Liver + Error(Treatment/Rat/Liver), rats)

using lme.
The code:

model1<- lme(Glycogen~Treatment, random = ~1|Rat/Liver, data=rats)
VarCorr(model1)

Variance     StdDev
Rat =       pdLogChol(1)
(Intercept) 20.6019981   4.538942
Liver =     pdLogChol(1)
(Intercept)  0.0540623   0.232513
Residual    42.4362241   6.514309

Does NOT give me the same variance componets I find in Crawley's book (page
371 onwards).
The code:

model2<- lme(Glycogen~Treatment, random = ~1|Treatment/Rat/Liver, data=rats)
VarCorr(model2)

Variance     StdDev
Treatment = pdLogChol(1)
(Intercept) 12.54061     3.541272
Rat =       pdLogChol(1)
(Intercept) 36.07900     6.006580
Liver =     pdLogChol(1)
(Intercept) 14.17434     3.764883
Residual    21.16227     4.600247

gets me very similar results (I would guess the differences are due to
rounding and the fact I am using R 1.6.2 and Crawley used S+).

My problem is: as *Treatment* is a fixed factor, why should I put it in
both the fixed term side and random terms side of my code to get the right
numbers? I fail to get this at all. Any elucidation would be appreciated.

Regards,

Federico Calboli

=========================

Federico C.F. Calboli

Department of Biology
University College London
Room 327
Darwin Building
Gower Street
London
WClE 6BT

Tel: (+44) 020 7679 4395
Fax (+44) 020 7679 7096
f.calboli at ucl.ac.uk

```