GLMER warning message
On 11-02-28 09:08 AM, Sophie Ledger wrote:
I am trying to investigate whether there are any trends in the visit rate of rabbits across 40 farms to the farms? building types over a year. For example, if they are visiting the feed store more in the spring than in summer. As it is a mixed effects model I have been trying to use a GLMER and loading the package lme4. The model structure so far is:
r1<-glmer(rabbitvisit~month+buildingtype+(1|farm),family=binomial
However, I when I run the model I get this message: Warning message: glm.fit: algorithm did not converge I also wish to nest ?date? as a random effect within farm to account for pseudoreplication but am not entirely sure how to proceed without understanding this warning message.
Very wise.
I have checked that rabbitvisit (presence/absence data) is read in as a factor and checked the raw data for any obvious errors. I tried to see if I could get the model to work with just month as a fixed effect and got a different warning message:
r3<-glmer(rabbitvisit~month+(1|farm),family=binomial) Warning message: In mer_finalize(ans) : false convergence (8)
I think (am not sure) that the "algorithm did not converge" warning is less serious that the "false convergence" warning (the former just indicates that there was a problem somewhere along the way, while the latter indicates a [possible] problem with the final model fit). How big is your data set (total observations, total number of visits, number of farms, number of months)? Are there months or farms or month/farm combinations that have exactly zero visits (or that are visited in every observation? Ben Bolker