Short Note
Test case for PEF estimation with sparse data

Jon Claerbout

claerbout@stanford.edu

Are PEFs going to be a useful tool on very sparse data sets? GEE disappoints. It throws away regression equations that depend on missing data. With sparse data, we have none left.

We need a test case. I offer one here. A test case with some realistic aspects. We should be able to get perfect results while nobody else does. And they'll all have to admit it :-)