st: margins with 3-way interaction

[Theocharis Kromydas <theocharis.kromydas@strath.ac.uk>]

Dear Statalisters

I am running a logistic regression like that  logistic empstatus year##country##c.edyears, where empstatus is employment status and edyears represents years of education.

I am running the margins command as margins year#country first but I am getting different results compared with margins, over(idround2 statelog),different with margins idround2, over(statelog) and again different results with margins statelog, over(idround2). Ok, differences are not big but I cannot really get it why they exist so I am confused which of these command is the right one. All I want is to plot the predicted empstatus over year, then over countries, then over years and countries and then over years, countries and years of education.

The problem is that my data is big and I don't have a good computer and fast version of stata to run it quickly. So only these commands can take up to half a day to be run and I also need to include control variables after and re run the model and margins command.

Any recommendations for this would be very much appreciated, as I am struggling with this lately not being sure what sort of results I should be getting from margins command.

Thanks,
Theocharis Kromydas 







*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/faqs/resources/statalist-faq/
*   http://www.ats.ucla.edu/stat/stata/