Generating Random Binary Deviates Having Fixed Marginal Distributions and Specified Degrees of Association

Abstract

Models for the analysis of correlated binary data have attracted considerable interest in the last few years. Parameter estimation in such models is based on asymptotic theory, and the small-sample properties of estimators must be studied by simulation. In this article we discuss several flexible methods for simulating random binary sequences with fixed marginal distributions and specified degrees of association between the variables. This last phrase is interpreted in two different ways and simulation methods appropriate to both are described.

Key Words:

• Copula
• Generalized estimating equation
• Linear programming