Goodness of fit measure for xtologit

[Sam McCaw]

Hello all,

I’m working with the xtologit model after quite a while and seem to have forgotten many of the modeling aspects. I ran a xtologit model and received a sigma_u value of higher than 8 and a rho somewhere around 89%. In general, does this mean that most of the variation in my data comes from between group variation? The LR test with a P >=chibar2 = 0.000 indicates that my panel model is better than the regular ologit model. Is this determined by the higher value of sigma_u meaning that there is enough variability between groups to use the xtologit model?

My second question is, what is the measure of goodness of fit for xtologit models and how do I compare goodness of fit between two xtologit models?

Finally, how does goodness of model fit impact the probability estimates in the panel ordered logit model?

Many many thanks.

SAM

[Carlo Lazzaro]

Sam:
1) your interpretation about the P>=chibar=0.000 is correct.
2) as far as I know, there's no such a test for -xtologit-.

[Sam McCaw]

Many thanks! Looking at various papers and stata manuals I could not find one either. In that case how do we know whether a model is a “good model” or if I should look for alternative specifications. Does the high value of sigma u represent a positive sign for the appropriateness of the model? For an xtologit what kind of diagnostics should I be looking at? Thanks!

[Carlo Lazzaro]

Sam:
1) a possible approach is to do our best to give a fair and true view of the data generating process (as it should be always the case).
2) see -help xtologit postestimation-.

[Sam McCaw]

Thanks again. I’ll do just that.

[Sam McCaw]

Thanks again. I’ll do just that.