Testing for pairwise independence.

This article presents a method for testing the hypothesis of mutual pairwise independence of k events. The method, which is based on the weighted least squares approach (Grizzle, Starmer, and Koch, 1969, Biometrics 25, 489-504) can be generalized to two types of incomplete data: the multiple-recapture census, where one of the cells of the corresponding 2k contingency table cannot be observed, and situations allowing an "unknown" response to the question designed to determine whether an event has occurred.