Georgia asks:

I’m examining how party level factors shape individuals’ attitudes using survey research data. I want to control for country level factors or at least allow countries to have random effects in the model. Individuals are within parties which are within countries, and each individual is only in one party. So, it’s a three stage hierarchical model. Within each party I have between 19-500 individuals and I have 50 parties. So, when I do the analysis without country characterisitcs – just looking at how party level factors shape individuals’ attitudes – I’m in good shape. However, when I add for country I have only 2-5 parties within each country. This is a problem, right?
I have been using a model with with country level random compenents and I’ve attached a file with the full model written out. But I think this may not be the right thing to do b/c I have so few parties within each country. Can you tell me if this is a problem and give me any advice on what I should do to control for country level effects?

My reply: 2-5 parties within each country is more than enough. See here for more discussion of this point.