Most of the analysis we have done so far assumes that cases were selected via simple random sampling – the equivalent of drawing names out of a hat. In reality, sampling schemes are often much more complicated than that. Survey data (I am quoting a lot from the Stata Manual here) are characterized by.