A REMARK ON ALGORITHM-643 - FEXACT - AN ALGORITHM FOR PERFORMING FISHERS EXACT TEST IN R X C CONTINGENCY-TABLES

The network algorithm of Mehta and Patel [1986] is currently the best general algorithm for computing exact probabilities in r x c contingen cy tables with fixed marginals. Given here are some improvements to th e network algorithm which speed its computational performance; and thu s increases the size of problems which can be handled. The new code al so eliminates some programming restrictions in the old code, implement s the ''hybrid'' algorithm of Mehta and Patel [1986a], and demonstrate s that the exact path length bounds of Joe [1988] should always be use d in place of the approximate bounds of Mehta and Patel [1986]. The ne w code can be much faster than the old code in some problems.