A program for the analysis of incomplete two-way cross-classifications.

Given (1) the size (R X C) of an incomplete two-way contingency table, (2) the apriori cell probabilities, lambda ij, (i = 1, 2, ..., R; j =1, 2, ..., C), the program produces 5 arrays of smoothed cell counts after adjusting existing zeros to certain values according to lambda ij and a weighting factor, K, which is also computed together with the value of a risk function for each array. After selecting one of those arrays according to the value of the risk function, then the program produces estimates for parameters of the saturated, independence and other hierarchial log-linear models in addition to the likelihood ratio for testing their goodness of fit.