The Use Of An EM Algorithm For Parameter Estimation From Sparse Data

Leon Aarons

Pharmacy Department, University of Manchester, Manchester 11139PL U.K.

A program was written implementing the EM algorithm of Dempster and Laird. Essentially the EM algorithm separates the structural estimation problem from the variance parameter estimation. Conditional (empirical Bayes) a posteriori individual parameter estimates are readily obtained. Results of applications to sparse data sets and covariate analyses will be presented to illustrate the technique.

Reference: PAGE 2 () Abstr 906 [www.page-meeting.org/?abstract=906]

Poster: oral presentation