Testing for ordered trends of binary responses between contingency tables

Abstract

In this article, likelihood ratio tests (LRTs) are developed for detecting that stochastic trends of binary responses are ordered between 2 × k contingency tables. We provide a simple iterative algorithm for the maximum likelihood estimators under the order restriction and construct the LRTs using those estimators. All the distributional results of these tests are based on the large sampling theory. The finite-sample behaviors of these tests are investigated through a simulation study. As an illustration of these tests, we analyze a set of data on wheeziness of smoking coalminers. © 2002 Elsevier Science (USA).