iteration limit reached without convergence

Dear mixed modellers,

 I am running some mixed-effects logit models that seem to fail to converge,
 as evidenced by the warning message:

 In mer_finalize(ans, verbose) :
 iteration limit reached without convergence (9)

Typically that is a symptom of a model that is over-specified.  Did
you try setting verbose = TRUE and checking what was happening to the
parameter estimates as they went through the iterations?  I mention
this because there is a tendency to specify models with every possible
covariate in the fixed effects and optimization of such models is
often difficult.  I don't want to make assumptions without having seen
the model and the data to which you are fitting it but my first
approach would be to simplify the model.

I suppose I should activate options in the development version of lme4
to set the maximum number of iterations.  One of the problems with
doing so is that it is not just a matter of the maximum number of
iterations.  One also needs to set the maximum number of function
evaluations.  If there are many parameters being optimized
simultaneously then the number of function evaluations can be much
larger than the number of iterations.

 In the CRAN version of lme4, it is possible to specify the maximum number
 of iterations used by nlminb while estimating model parameters, by passing
 the parameter control=list(msMaxIter=N) in to lmer. However, this doesn't
 seem to work with the R-Forge version. Is there a way of controlling the
 iteration limit in the R-Forge version that I'm overlooking?

 Much obliged,
 Gabe Doyle

 _______________________________________________
 R-sig-mixed-models at r-project.org mailing list
 https://stat.ethz.ch/mailman/listinfo/r-sig-mixed-models