[R-sig-ME] The issue of increasing maximum number of iterations

[Ben Bolker]

Martin Stjernman <Martin.Stjernman at ...> writes:

> 
> Dear listmembers!
> 
> I have also had trouble with the maxIter specification in (g)lmer. 
> I've tried to increase the number of
> iterations (to 1000) via the control statement in lmer 
> but it still stops at 300 (the default).
> We are now a number of users having problem with this but 
> I haven't seen any solution to the problem on this
> list and I would really appreciate any hint on what I am doing wrong.
> I am running on a Vista 64-bit machine and my model specification and
>  session info follows below. Please let
> me know if any other information needs to be attached.

  Looking at the source code in lmer.R, this seems like
a moderately obvious hole: on or about line 714 (I may be
off by a couple of lines) we have

    if (missing(verbose)) verbose <- cv$msVerbose
### FIXME: issue a warning if the model argument is FALSE.  It is ignored. 

 adding the lines

    FL$dims["mxit"] <- cv$maxIter
    FL$dims["mxfn"] <- cv$maxFN

  appears to work: running

 (gm1 <- glmer(cbind(incidence, size - incidence) ~ period + 
   (1 | herd), family=binomial, data=cbpp, control=list(maxIter=1000)))

produces a model with the correct iteration number listed in its
guts.

  The next question is whether you or someone at your institution
is capable of modifying lmer.R appropriately and rebuilding the
package ...