Gang Wrote:
-------------------------------------------------------
> > lme(values ~ time*treatment +
> ratings,data=lme_data,random = ~1 + ratings |
> subj)
>
> > I get the following error message:
> > nlminb problem, convergence error code = 1
> > message = iteration limit reached without
> convergence (10)
>
> This is basically what 3dLME does except for a
> couple of differences (see below). Basically it
> failed to converge at that voxel for some
> numerical reason. Try a different voxel, and see
> if it converges.
>
Gang, this a follow-up about failing to converge and how 3dLME handles it.
When I run the following command in R for one voxel:
lme(values ~ time*treatment + ratings,data=lme_data,random = ~1 + ratings | subj)
I get the failing to converge error message.
Yet, for 3dLME, I get results for that voxel, MainF's, contrasts, etc.
How is 3dLME handling failing to converge issues?
I did notice if I take ratings out as a random factor:
lme(values ~ time*treatment + ratings,data=lme_data,random = ~1 | subj)
The above command in R works. Oddly enough, the mainF, interaction, and intercept results without treating ratings as a random effect in the lme command match my results from 3dLME while treating ratings as a random effect.
Why is this?
Another question:
How does 3dLME handle missing voxels? For example, if you have 10 subjects, and in one area the brain a given voxel is present for 9 of the 10 subjects? Is the 3dLME analyzed for just the 9 subjects or is it treated as NA? For one of our voxels, we've run into this issue where there are NA values for some subjects at a given level of treatment and/or time. We received the following error message trying to run the lme command above with a data set containing NA's:
Error in na.fail.default(list(values = c(NA, NA, NA, NA, NA, NA, NA, NA, :
missing values in object
Thanks,
Michael