[R-sig-ME] Predict method for lmer

[Ben Bolker]

On 13-11-29 12:07 PM, Nathan E. Rutenbeck wrote:
> 
>>> Hi all,
>>>
>>> I am struggling to use predict.merMod to generate predictions from a
>>> relatively simple mixed model.
>>   [snip snip snip]
>>
>>> Example code follows:
>>>
>>> mer0 <- lmer ( top ~ 1 + (1|stand), data=data)
>>> mer0.1 <- update( mer0, .~. + (1|species))
>>   [more snippage]
>>  
>>> mer2 <- update(mer1, .~. + (DBH|Species)
>>> predict(mer2, newdata=data, allow.new.levels=T) # This fails with the
>>> following error:
>>>
>>> Error in t(.Call(Csparse_dense_crossprod, y, t(x))) :
>>>     error in evaluating the argument 'x' in selecting a
>>>   method for function
>>> 't': Error: Cholmod error 'X and/or
>>> Y have wrong dimensions' at file ../MatrixOps/cholmod_sdmult.c, line 90
>>>
>>> It seems that I am failing to provide the proper matrix format for the
>>> newdata argument, but I don't necessarily see how to fix it.
>>>
>>> Thanks in advance,
>>>
>>> -Nathan Rutenbeck
>>>
>>    This is probably a bug (i.e. not your fault), but possibly one that
>> has
>> been fixed more recently. Can you please:
>>
>> * try installing the development version, either from Github or via
>> install.packages("lme4",repos="http://lme4.r-forge.r-project.org/repos")
>> and see if that helps?
>> * send me a _small_ reproducible example?
>>
>>    Ben Bolker
> Hi Ben,
> 
> Thanks for the tip. I can report back that the development version of
> lme4 solved all my woes.
> 
> -Nathan

  Thanks for reporting back, and thank goodness your problem is already
resolved in the development version -- we're deep in petty bugs here at
lme4 central, we don't need any more (although that shouldn't discourage
people from reporting them if they're out there!)

  Ben Bolker