Mixed Effects Models

[D Stroup]

Dear STATALIST users, I am interested in your views on analysis of a mixed effects model. The model is a linear regression predicting a continuous variable. The predictor variables are in 2 categoris: some at the individual level and some at an aggregate level. For example, the objective is to predict selling price of a property. We have some property-level variables (lot size, e.g.) and some variables are available at only an aggregate level (e.g., zip code). This is not the usual xtmixed set up. Does anyone have experience in this sort of analysis? Thanks.

[Timothy Mak]

I'd imagine that this IS the usual -xtmixed- set up. A typical model might go:

xtmixed sellingprice propertyvar1 propertyvar2 areavar1 areavar2 || zip:

Note that you normally include all predictor variables in the fe part of the model, unless you want their effects to vary by area.

Tim