ON THE ESTIMATION OF EMPTY CELL PROBABILITIES IN A CONTINGENCY TABLE

Abstract

In this paper, an Independent Binary Model (IBM) is proposed. It is aimed at estimating cell probabilities in an r x c contingency table when some of the cells have zero count. Existing methods in this situation are either subjective or based on arbitrary decision of the researcher. The IBM is applied to sets of simulated data for various combinations of categorical variables. It is pointed out that the IBM could be an alternative for such situations especially when the result is needed for further analysis.