This report is released to inform interested parties of ongoing research and to encourage discussion. Any views ex- pressed on technical issues are those of the author and not necessarily those of the U.S. Census Bureau.
Michael Moldoff · 2009
The Expectation-Maximization (EM) Algorithm is sometimes used to impute incomplete or missing data. The algorithm utilizes maximum likelihood estimators (MLEs) to produce estimates. However, programming this entire method in SAS® can be tedious and tricky when many calculations and probabilities are involved, or because of the complexity of a particular model’s MLE. SAS has an EM Algorithm option in the procedure PROC MI. SAS imputes values for missing observations for a numeric variable using the other existing data. However, when looking at examples on the web, I could not find a method in SAS that used the EM Algorithm to impute categorical variables. To better illustrate the difference, here is an example of the data that I found the SAS EM Algorithm ready to handle: data EMdata;