A Pseudo-Exact Sign Test for Cluster Correlated Data* Michael D. Ernst Indiana University-Purdue University Indianapolis The difficulties of obtaining an exact permutation distribution for the sign test in cluster correlated data are discussed. A pseudo-permutation distribution of the sign statistic is proposed and is evaluated for its control of the Type I error rate. Its power is also compared to available methods. *Based on joint work with Denis Larocque, École des Hautes Études Commerciales de Montréal