[R] Problem with nlme or glmmPQL (MASS)

[Prof Brian Ripley]

glmmPQL is not documented to work with groupedData objects, so that's your 
mistake.

See ?glmmPQL and supply a valid formula for random.

On Mon, 14 Apr 2003, kjetil brinchmann halvorsen wrote:

> I am encountering the following problem, in a multilevel analysis, 
> using glmmPQL from MASS. This occurs with bothj rw1062 and r-devel, 
> respectively with nlme versions 3.1-38 and 3.1-39 (windows XP).
> 
> >  S817.mod1 <- glmmPQL( S817 ~ MIEMBROScat+S901+S902A+S923+URBRUR+REGION+
> +                  
> S102+S103+S106A+S108+S110A+S109A+S202+S401+S557A+S557B+
> +                  YHOGFcat, data=Hogares, random= ~1, 
> family=binomial, 
> +                  na.action=na.omit)
> iteration 1 
> Error in getGroups.data.frame(dataMix, groups) : 
>         Invalid formula for groups
> > getGroupsFormula(Hogares)
> ~ID08
> 
> Hogares is a groupedData object. The only non-default thing I have 
> done is to use order.groups=FALSE in groupedData, since the response 
> is a factor and the max ordering function does not apply. 
> 
> What is wring here?

-- 
Brian D. Ripley,                  ripley at stats.ox.ac.uk
Professor of Applied Statistics,  http://www.stats.ox.ac.uk/~ripley/
University of Oxford,             Tel:  +44 1865 272861 (self)
1 South Parks Road,                     +44 1865 272866 (PA)
Oxford OX1 3TG, UK                Fax:  +44 1865 272595