Hello,

I have a model that is of the form:
Code:
reg y x i.z i.year
which runs without any errors.

I am wondering why STATA does not drop any of the regressors due to collinearity?

where:
  • z is a categorical variable with ~200 levels
  • x is continuous and by definition, constant within z and year; that is, z and year perfectly predict x
  • y is continuous
Since there is no variation of x within z and year, where is the estimation of the effect of x on y (conditional on z and year) coming from in my model?

This really puzzles me.

Thanks,
Karl