[R-sig-ME] Warning message

[Ben Bolker]

As previously suggested/discussed elsewhere, especially under the
?convergence manual page, this is probably a false positive. The gold
standard is to choose a different optimizer to fit your model with and
convince yourself that the results are sufficiently close across
optimizers that the warning is indeed a false positive.  (The chances
of two separate optimizers misconverging to the same point are small
... they might or might not reach different local optima, but that's
not the kind of convergence failure that the tests are able to detect
...)

  Have you been warned about the dangers of stepwise model selection
<http://www.stata.com/support/faqs/statistics/stepwise-regression-problems/>
?


On Wed, Dec 9, 2015 at 5:32 AM,  <katrin.luder at iee.unibe.ch> wrote:
> Hello
>
> I'm currently working on my master thesis and i'm using the lme4 package to do glmer models. To find the best model, i'm using the forward selection approach. After putting the former explanatory variables in, everything works fine, but after a while this error message is showing up:
>
> Warning message:
> In checkConv(attr(opt, "derivs"), opt$par, ctrl = control$checkConv,  :
>   Model failed to converge with max|grad| = 0.00377805 (tol = 0.001, component 2)
>
> I have found different reasons in Google but nothing really fitted my problem. At some forum, it was written that i can also just skip this error message and go on with what i was doing. But the more explanatory variable i put in, the more often this error message occurs.
>
> I would be very happy if you could help me out with this problem. I already asked several ecologist and statisticians but no one was able to help me.
>
> Your sincerely,
> Katrin Luder
>
>         [[alternative HTML version deleted]]
>
> _______________________________________________
> R-sig-mixed-models at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-sig-mixed-models