R help - Apr 2009 - refit with binomial model (lme4)

Dear R users,

I'm trying to use function 'refit' from lme4
and I get this error that I can't understand:
> refit(dolo4.model4,cbind(uu,50-uu))
Error in function (classes, fdef, mtable)  : 
  unable to find an inherited method for function "refit", for
signature
"mer", "matrix"

if I try:
> refit(dolo4.model4,uu)
Error in asMethod(object) : matrix is not symmetric [1,2]

I get this error message that I can no more
understand but which suggests that refit expects
two columns.


the initial model was:
> dolo4.model4 at call
glmer(formula = cbind(sortis, restes) ~ mean.co2 + (1 | sujet), 
    data = dollo4.df, family = binomial)



R version 2.9.0 (2009-04-17)

and

Package: lme4
Version: 0.999375-28
Date: 2008-12-13

thank you in advance

e.e.

This is related to using the matrix form of the response for a
binomial glmm.  The refit method for a model fit by lmer is based on a
numeric vector response.

Is it possible to use the expanded form (i.e. a vector of 0/1 values)
of the responses instead of the matrix form?

On Mon, Apr 27, 2009 at 7:20 AM, Eric Elguero <Eric.Elguero at mpl.ird.fr>
wrote:
> Dear R users,
>
> I'm trying to use function 'refit' from lme4
> and I get this error that I can't understand:
>
>> refit(dolo4.model4,cbind(uu,50-uu))
> Error in function (classes, fdef, mtable) ?:
> ?unable to find an inherited method for function "refit", for
signature
> "mer", "matrix"
>
> if I try:
>
>> refit(dolo4.model4,uu)
> Error in asMethod(object) : matrix is not symmetric [1,2]
>
> I get this error message that I can no more
> understand but which suggests that refit expects
> two columns.
>
>
> the initial model was:
>
>> dolo4.model4 at call
> glmer(formula = cbind(sortis, restes) ~ mean.co2 + (1 | sujet),
> ? ?data = dollo4.df, family = binomial)
>
>
>
> R version 2.9.0 (2009-04-17)
>
> and
>
> Package: lme4
> Version: 0.999375-28
> Date: 2008-12-13
>
> thank you in advance
>
> e.e.
>
> ______________________________________________
> R-help at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide
http://www.R-project.org/posting-guide.html
> and provide commented, minimal, self-contained, reproducible code.
>

Thanks for sending that, Ben.  I have incorporated a slightly modified
version, with attribution, in the lme4 sources.

On Tue, Apr 28, 2009 at 8:03 AM, Ben Bolker <bolker at ufl.edu> wrote:
> I think the following works, but use with caution -- I just
> wrote it last night.
>
> setMethod("refit", signature(object = "mer", newresp =
"matrix"),
> ? ? ? ? ?function(object, newresp, ...)
> ? ? ?{
> ? ? ? ?## newresp <- as.double(newresp[!is.na(newresp)])
> ? ? ? ?wts <- rowSums(newresp)
> ? ? ? ?newresp <- newresp[,1]/wts
> ? ? ? ?stopifnot(length(newresp) == object at dims["n"])
> ? ? ? ?object at y <- newresp
> ? ? ? ?object at pWt <- wts
> ? ? ? ?lme4:::mer_finalize(object)
> ? ? ?})