[R-sig-ME] Error when using models with lmer

[Douglas Bates]

On Wed, Feb 8, 2012 at 2:11 PM, Douglas Bates <bates at stat.wisc.edu> wrote:
> On Wed, Feb 8, 2012 at 12:52 PM, Olga Vitek <ovitek at stat.purdue.edu> wrote:
>> Dear Dr. Bates,
>>
>> May I ask you for a clarification? The two problematic functions are indeed from the package gmodels
>>
>> ci(fit4)                                         # 'ci' stands for confidence intervals of the fixed effects
>> estimable(fit4, c(0, -1, 0, 0))       # produces estimates of SE of the linear combination of the fixed effects
>>
>> The difficulty is that there was no change in the version of gmodels before or after the problem occurred. The code worked with lme4_0.999375-40 and R2.13.1, but not with lme4_0.999375-42 and R2.14.1.
>
> You need to go through some hidden functions to find out where the
> problem lies.  Eventually it boils down to a hidden function called
> gmodels:::est.mer which calls mcmcsamp and that function is no longer
> available.

Again, I misspoke.  It's just not my day.  The mcmcsamp function is
still available in the released version of lme4.  Actually the code in
gmodels is trying to convert the merMCMC object to a matrix and
somehow isn't seeing the necessary method for as.matrix.  It will have
to do with namespaces and the best way to solve that would be in the
NAMESPACE file for the gmodels package.

>> Could you kindly let me know if there were changes in the structure of the class 'mer' since lme4_0.999375-40? Alternatively, could you recommend other options for calculating standard errors of linear combinations of fixed effects?
>>
>> Many thanks in advance
>> Olga Vitek
>>
>>
>>
>>
>>
>> On Feb 8, 2012, at 11:51 AM, Douglas Bates wrote:
>>
>>> On Tue, Feb 7, 2012 at 2:03 PM, Olga Vitek <ovitek at stat.purdue.edu> wrote:
>>>> Hello,
>>>>
>>>> I am executing the following code
>>>>
>>>>> library("lme4")
>>>>> library(faraway)
>>>>> data(penicillin)
>>>>> fit4 <- lmer(yield ~ treat + (1|blend), penicillin)
>>>>
>>>> and the lines below generate error messages
>>>>
>>>>> library(gmodels)
>>>>> ci(fit4)
>>>
>>> Yes.  Why would you want to try to create a vector from an S4 fitted
>>> model object?
>>>
>>> It almost never makes sense to use
>>>
>>> c(foo)
>>>
>>> Most of the time when people use that idiom what they really mean is
>>>
>>> foo
>>>
>>> (I say "almost never" because in the old days many of us would use
>>> c(myMatrix) to create a vector from a matrix.  The preferred form is
>>> now as.vector(myMatrix).)
>>>
>>>> Error in as.vector(data) :
>>>>  no method for coercing this S4 class to a vector
>>>>
>>>>> contrast.function <- c(0, -1, 0, 0)
>>>>> test.result <- estimable(fit4, contrast.function)
>>>> Error in as.vector(data) :
>>>
>>> I assume that the "estimable" function is in the gmodels package, so
>>> you should contact the author of that package about this.
>>>
>>>>  no method for coercing this S4 class to a vector
>>>>
>>>>
>>>> This worked with with lme4_0.999375-40 and R version 2.13.1, but not with lme4_0.999375-42 and R version 2.14.1.
>>>>
>>>> Would you have suggestions on how to make this work?
>>>> Thank you in advance
>>>> Olga Vitek
>>>>
>>>>> sessionInfo()
>>>> R version 2.14.1 (2011-12-22)
>>>> Platform: i386-apple-darwin9.8.0/i386 (32-bit)
>>>>
>>>> locale:
>>>> [1] en_US.UTF-8/en_US.UTF-8/en_US.UTF-8/C/en_US.UTF-8/en_US.UTF-8
>>>>
>>>> attached base packages:
>>>> [1] stats     graphics  grDevices utils     datasets  methods   base
>>>>
>>>> other attached packages:
>>>> [1] gmodels_2.15.1   faraway_1.0.5    lme4_0.999375-42 Matrix_1.0-3     lattice_0.20-0
>>>>
>>>> loaded via a namespace (and not attached):
>>>> [1] gdata_2.8.2   grid_2.14.1   gtools_2.6.2  MASS_7.3-16   nlme_3.1-102  stats4_2.14.1 tools_2.14.1
>>>> _______________________________________________
>>>> R-sig-mixed-models at r-project.org mailing list
>>>> https://stat.ethz.ch/mailman/listinfo/r-sig-mixed-models
>>