[R-sig-ME] question about lme4

[Ben Bolker]

On 14-07-18 10:32 AM, Leithen wrote:
> Hi Ben,
> Thank you for this. I hadn't seen it. Does this apply to 1.1-7 as well?
> --L

  [cc'ing r-sig-mixed-models]

  No, it doesn't; in 1.1-7 we have changed the tests so that it is
much more likely that reported convergence failures represent a real
problem rather than a false positive. They *might* still be false
positives, but for now we'd rather worry people while we figure out
how to tweak the messages so they're slightly less sensitive (but
still don't miss things that are real problems).

  Ben Bolker






>
>
> On Thu, Jul 17, 2014 at 7:47 PM, Ben Bolker <bbolker at gmail.com> wrote:
>
> On 14-07-17 07:49 PM, Leithen wrote:
>>>> Hi Ben,
>>>>
>>>> I am a post-doc at UC Berkeley (ecology and evolutionary biology)
>>>> and was hoping I could ask you a quick question. I use lme4 a fair
>>>> bit for my analyses and I recently upgraded to the latest version
>>>> of R and lme4. I am now finding that analyses which previously ran
>>>> without warnings are giving me convergence warnings when they
>>>> didn't before. E.g., this type of error:
>>>>
>>>> Warning message: In checkConv(attr(opt, "derivs"), opt$par, ctrl =
>>>> control$checkConv,  : Model failed to converge with max|grad| =
>>>> 0.0119784 (tol = 0.001, component 3)
>>>>
>>>> I'm just wondering if you have an idea why? Have you updated lme4
>>>> so that it now prints out warnings more readily?
>>>>
>>>> Thank you for your help. I am at a bit of a loss for ideas why this
>>>> is happening. Cheers, --Leithen
>
>   Are you using version 1.1-6?  Have you looked at
> https://github.com/lme4/lme4/blob/master/README.md ?
>
>   cheers
>     Ben Bolker