[R-sig-ME] lmer()----mer_finalize(ans) : singular convergence (7)

[Dennis Murphy]

>From Ben Bolker's GLMM wiki:  http://glmm.wikidot.com/faq


Gamma GLMMs?

While one (well, ok I) would naively think that GLMMs with Gamma
distributions would be just as easy (or hard) as any other sort of
GLMMs, it seems that they are in fact harder to implement (or at least
not yet thoroughly/stably working in lme4). Basic simulated examples
of Gamma GLMMs can fail in lme4 despite analogous problems with
Poisson, binomial, etc. distributions. Solutions:

    * the default inverse link seems particularly problematic; try
other links if that is possible/makes sense
    * consider whether a lognormal model (i.e. a regular LMM on logged
data) would work/makes sense
    * glmmPQL
    * other packages (WinBUGS) etc. ?
    * gamlssNP?

HTH,
Dennis

On Sun, Jun 26, 2011 at 7:21 AM, rajibul islam <rimian85 at hotmail.com> wrote:
>
> Dear:
>
> I am working on gamma mixed model. I have used...... lmer().... to model the data set. Here is the short part of the data.....
>
>      obs treat*  id time response timesq responsebin
> 1       1     0   2    0      4.0      0           0
> 2       2     0   2    1      6.0      1           0
> 3       3     0   2    2      7.0      4           0
> 4       4     0   2    3      9.0      9           0
> 5       5     0   2    6     13.0     36           0
> 8       8     0   3    1      1.0      1           0
> 9       9     0   3    2      2.0      4           0
> 10     10     0   3    3      3.0      9           0
> 11     11     0   3    6      9.0     36           0
> 12     12     0   3    9      4.0     81           0
> 13     13     0   3   12      2.0    144           0
>
>
> *treatment has two groups......... 0 and 1
>
> I used the model ....
>
> m1.3.cg2<-lmer(response~time+timesq+factor(treat)+(1|time)+(1|id)+(time|id),data=dcg,
> family=Gamma(link = "identity"))
>
> And it shows the warning message like........
>
> Warning message:
> In mer_finalize(ans) : singular convergence (7)
>
> Moreover, the result it shows is quite confusing due to large estimated value of fixed effects..................
>
> My concern is whether I should keep the model with such warning message or not?
>
> And what this warning message indicates?
>
> Any comments or suggestion would be highly appreciable.
>
> Thank you.
>
> Rajibul Mian, Grad student, UNB, Canada
>
>
>
>        [[alternative HTML version deleted]]
>
> _______________________________________________
> R-sig-mixed-models at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-sig-mixed-models
>