[R-sig-ME] Bglmer convergence warnings (covariance structure)

[Ben Bolker]

 I think if you'd failed to specify the covariance matrix correctly that
you would simply have gotten an error.
  I'm doing some test runs on simulated data. I did get convergence
warnings, but the results look reasonable when compared across widely
different fitting platforms (lme4, blme, glmmTMB, brms):

http://bbolker.github.io/mixedmodels-misc/notes/bglmer_cmp.html

code (Rmd file etc.) is in the corresponding github repo

  cheers
   Ben Bolker



On 17-06-12 06:45 AM, D.J. Damen wrote:
> Dear all,
> 
> I am trying to fit a binomial model with three factors using the bglmer package in R. My model includes a complex 3x2x2 design; factor c.con.tr has three levels, factor c.type.tr two, and factor c.diff.tr has also two levels. When I try to fit a random-intercept only model, the model produces the following warning:
> 
> Warning message:
> In get("checkConv", lme4Namespace)(attr(opt, "derivs"), opt$par,  :
>   Model failed to converge with max|grad| = 2.02439 (tol = 0.001, component 1)
> 
> I figured that I did not specify my covariance structure correctly, but I am at a loss as to how I should change it. The structure of my model is as follows:
> 
> riobglmer <- bglmer(contrast~c.con.tr*c.type.tr*c.diff.tr+(1|id)+(1|item.new), data=mydata, family=binomial (link='logit'), fixef.prior= normal(cov = diag(9,12)))
> 
> To me, (cov = diag(9,12)), seems to be correct.. Am I missing something important? Suggestions and/or remarks are more than welcome.
> 
> Thank you very much in advance.
> 
> Best regards,
> 
> Debby Damen
> PhD Student
> 
> Department of Communication and Information Science
> 
> Tilburg University
> Warandelaan 2, room D410
> 5037 AB Tilburg
> 
> T. +31 13 466  8245
> M. d.j.damen at uvt.nl<mailto:d.j.damen at uvt.nl>
> 
> 
> 
> 	[[alternative HTML version deleted]]
> 
> _______________________________________________
> R-sig-mixed-models at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-sig-mixed-models
>