[R-sig-ME] Error in .local(object, n, verbose, ...) : Update not yet written

[Douglas Bates]

On Wed, Apr 27, 2011 at 8:50 AM, Gregory Gilbert
<g.eastham.gilbert at gmail.com> wrote:
> Hello All,
>
> I am running the following code under R.2.11.1 on a Dell PowerEdge 2900 III
> running Windows 2003 Server 64-bit OS (Release 2 build 3790) with 48GB RAM.
>
> When I try an get HPD intervals for my model using mcmcsamp() I get the
> following error: "Error in .local(object, n, verbose, ...) : Update not yet
> written"
>
> I have searched the R-help & R-sig-mixed archives; I have googled the
> specific error (and found some threads in R-help dating to 2008-2009
> suggesting the author of the code put this message in mcmcsamp to remind him
> to finish coding a particular part of mcmcsamp); I have examined the lme4
> documentation (and mcmcsamp); I have read the R FAQ; and I have consulted
> with my boss PhD biostatistician (he suggested I make a post). As mentioned,
> I did run across a number of conversations dating to 2008-2009 regarding
> this error, but did not see a suggestion. I did see something to the effect
> that it was to remind the programmer he needed to finish some coding. Was
> the code ever finished/updated? Is the issue still outstanding? I have tried
> installing lm4a from R-forge and ran into the following error:

You didn't tell us about your model.  What is the model?  Is it a
linear mixed-effects model, a generalized linear mixed-effects model,
...?  Does it have only scalar random effects or are there
vector-valued random effects?
>> install.packages("lma4", repos="http://R-Forge.R-project.org")
> Warning: unable to access index for repository
> http://R-Forge.R-project.org/bin/windows64/contrib/2.11
> Warning message:
> In getDependencies(pkgs, dependencies, available, lib) :
>  package ‘lma4’ is not available
>
> when I researched this error it was suggested I install "Rccp" from R-forge.
> I tried this and received the following error:
>
>> install.packages("Rcpp", repos="http://R-Forge.R-project.org")
> Warning: unable to access index for repository
> http://R-Forge.R-project.org/bin/windows64/contrib/2.11
> Warning message:
> In getDependencies(pkgs, dependencies, available, lib) :
>  package ‘Rcpp’ is not available

The problem with installing the Rcpp package is that recent versions
depend on an R version >= 2.12.0 and you are running version 2.11.
Similarly for the lme4 (which you misspelled as "lma4", although
perhaps you meant "lme4a").

First I would recommend installing R-2.13.0, the most recent version.
Then try installing lme4 from CRAN.  If you want to try lme4a then
install Rcpp, minqa and MatrixModels from CRAN and, finally,

install.packages("lme4a", repos="http://lme4.R-forge.R-project.org/repos")



> Where am I going wrong in the process? What have I missed? Is this a problem
> with my data? I have run a similar regression with continuous A1c data
> (range: 3 to 18) as the dependent variable and had no problems for all my
> data (n=700K observations and 5 time points). I also ran the data for 1%,
> 5%, 10%, and 25% random samples or the data, with a continuous dependent
> variable) and had no problems generating CIs. I am not wedded to mcmcsamp()
> & HPDinterval(), but I understand it is one of the more robust methods of
> generating CIs about beta for mixed effects models. I am happy to use any
> approach as long as statistically valid.
>
> Any status or updates would be greatly appreciated.
>
> Thanks in advance.
>
> Greg Gilbert
>
>
>> options(repos="http://lib.stat.cmu.edu/R/CRAN/")
>> myPackages <- c("lme4", "foreign", "Hmisc")
>> install.packages(myPackages, dependencies=TRUE)
>> update.packages(checkBuilt=TRUE, ask=FALSE)
>> library(foreign)
>> library(Hmisc)
>> library(lme4)
>>
>> cohort <- sasxport.get("//v07.med.va.gov/cha/TREP/___neworganization/workgroups/wg10_06_sampling_gebregziabher/data/PCT01SRS.xpt")
> Processing SAS dataset PCT01SRS          ..
>> ## time is 1 to 5
>> cohort$time.ctr.sq <-(cohort$time-3)*(cohort$time-3)
>> attach(cohort)
>> ## a1cge8 is 0, 1
>> ## id is 4-7 digit number
>> ## age is continuous 18-103
>> ## svcpct is continuous 0 to 100
>> ## all other variables are binary 0, 1
>> f  <- a1cge8 ~  (1|id)   + time       + time.ctr.sq     +
> +                 nhb      + hispanic   + other           +
> +                 male     + age        + mstat           +
> +                 svcpct   + urban      +
> +                 abuse    + anemia     + cancer          +
> +                 cbd      + chf        + cvd             +
> +                 depress  + htn        + hypothy         +
> +                 liver    + lung       + lytes           +
> +                 obese    + other.dz   + psych           +
> +                 pvd
>>
>> ## finite population correction
>> population <- 890394
>> sample     <-   9000
>> fpp        <- sqrt((population-sample)/(population-1))
>>
>> fit        <- lmer(f, family=binomial(link="logit"), data=cohort)
>>
>>    mcmc.ci <- mcmcsamp(fit, 1000)
> Error in .local(object, n, verbose, ...) : Update not yet written
>> ci         <- HPDinterval(mcmc.ci)
> Error: object 'mcmc.ci' not found
> Error in HPDinterval(mcmc.ci) :
>  error in evaluating the argument 'object' in selecting a method for
> function 'HPDinterval'
>>
>
>        [[alternative HTML version deleted]]
>
>
> _______________________________________________
> R-sig-mixed-models at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-sig-mixed-models
>
>