Daniel Ezra Johnson <johnson4 <at> babel.ling.upenn.edu> writes:
1) Yes, I have tweaked the data to show as clearly as I can that this is a bug, that a tiny change in initial conditions causes the collapse of a reasonable 'parameter' estimate.
I would not call this a bug, since this is related to data and not to the software. I might be wrong!
2) mcmcsamp() does not work (currently) for binomial fitted models.
Sorry, for wrong pointer. You could try with some other packages if they have support for binomial models with "random" effects. I would just try in BUGS --> take a look at R2WinBUGS or Brugs.
3) This is an issue of what happens when the sample is too small. For all larger data sets I have gotten a ranef variance between 0.05 and 1.00 or so. It makes no sense to say that as the data set gets smaller, the systematic variation between Items goes away. It doesn't, as I've shown. In the data
I believe that when data gets smaller such parameters are harder to estimate and you can easily get 0 as MLE.
above, certain Items were still 10+ times as likely (log-odds wise) to have Response==1 as others.
Gregor