CONSTRAINED EMPIRICAL BAYES ESTIMATES FOR CELL VALUES IN A 2-WAY TABLE

It is common to summarize the results of a forest inventory in two-way tables. Unfortunately, while the overall sample size may be large, the sample size for an individual cell in the table may be quite small. Thus the estimate may have a large standard error. We propose a simultaneous estimation method to reduce the variance and (or) mean squared error of individual cell estimates while retaining table additivity, i.e., preserving the observed row and column sums of the table.