Implements a method of iteratively collapsing the rows of a contingency table, two at a time, by selecting the pair of categories whose combination yields a new table with the smallest loss of chi-squared, as described by Greenacre, M.J. (1988) <doi:10.1007/BF01901670>. The result is compatible with the class of object returned by the 'stats' package's hclust() function and can be used similarly (plotted as a dendrogram, cut, etc.). Additional functions are provided for automatic cutting and diagnostic plotting.