Good evening,

I am running two regressions with gllamm, one for each of two control groups, with identical variables. While in one of the two models no matter what the order of the variables is, the results remain the same, as expected, that is not the case with the second one. I tried to find a possible culprit variable by following a stepwise-like process: adding an extra variable each time and re-running the regression in a different variables order. It initially turned out that my (ordered) income variable was causing the trouble, however when I iterated the rerunning process, this time adding income earlier, it was not causing a problem anymore; the problem showed up after I have included a few more variables. I also thought that the relationship between income and another independent variable might be causing the problem, but it seems that no specific variables or combination of variables are to be blamed for. When I run the same models with xtlogit (which does not have the pweight option), the problem does not appear. I dont know if it helps but the regression looks something like that:

gllamm WTP bid x1 ... xn, nolog i(id) link(logit) family(binomial) pw(wght)

Has anyone come across anything like this before? Any idea of where the problem could be coming from?

thanks in advance,
Marios