[R] anova grouping of factors in lme4 / lmer

[William Valdar]

Hi. I'm using lmer() from the lme4 package (version 0.8-3) and I can't get
anova() to group variables properly. I'm fitting the mixed model

  Response ~ Weight + Experimenter + (1|SUBJECT.NAME) + (1|Date.StudyDay)

where Weight is numeric and Experimenter is a factor, ie,

> str(data.df)
`data.frame':	4266 obs. of  5 variables:
  $ SUBJECT.NAME : Factor w/ 2133 levels "A0480","A0..",..: 1 1 2 2 ...
  $ Response     : num  -0.490  0.145  1.992 -0.391  0.917 ...
  $ Date.StudyDay: Factor w/ 161 levels "16","22","24",..: 24 24  ...
  $ Experimenter : Factor w/ 7 levels "amyt","carmena",..: 5 5 1 1 3 3 ...
  $ Weight       : num  25.3 25.3 16.7 16.7 28.2 28.2 27.1 27.1 ...

The model fits fine but when I call anova() to see the effect of the
Experimenter factor, I get a sequential anova table where factor
levels are fit one at a time, ie,

> anova(lmer(Response ~ Weight + Experimenter
+        + (1 | SUBJECT.NAME) + (1 | Date.StudyDay),
+        data=data.df))
Analysis of Variance Table
                      Df  Sum Sq Mean Sq  Denom F value    Pr(>F)
Weight                1    10.8    10.8 4242.0 14.4457 0.0001463 ***
Experimentercarmena   1     0.5     0.5 4242.0  0.6708 0.4128136
Experimentercnunez    1     4.8     4.8 4242.0  6.5042 0.0107969 *
Experimenterlsolberg  1     0.4     0.4 4242.0  0.5910 0.4420783
Experimenterpeterb    1 0.01430 0.01430 4242.0  0.0192 0.8898029
Experimenterpolinka   1     0.2     0.2 4242.0  0.2290 0.6322775
Experimenterstuart    1     0.9     0.9 4242.0  1.2505 0.2635116

... rather than what I want, which is a sequential anova table where the
levels are grouped, eg [using lme()],

> anova(lme(Response ~ Weight + Experimenter, random = ~ 1 | SUBJECT.NAME,
+       data=data.df))
Analysis of Variance Table
              Df Sum Sq Mean Sq  Denom F value    Pr(>F)
Weight        1    9.5     9.5 4242.0 12.7213 0.0003655 ***
Experimenter  6    8.5     1.4 4242.0  1.9067 0.0759496 .

Does anyone know how to make anova() group properly with lmer objects?

Many thanks,

William

=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=
Dr William Valdar               ++44 (0)1865 287 717
Wellcome Trust Centre           valdar at well.ox.ac.uk
for Human Genetics, Oxford      www.well.ox.ac.uk/~valdar