Models for Correlated Mixed Discrete and Continuous Random Variables in Clinical Trials

• Activity Number: 562
• Type: Contributed
• Date/Time: Wednesday, August 6, 2014 : 2:00 PM to 3:50 PM
• Sponsor: Biometrics Section
• Author(s): Sergei Leonov*+ and Bahjat Qaqish
• Companies: AstraZeneca and University of North Carolina at Chapel Hill
• Keywords: marginal distribution ; multivariate distribution ; extreme correlation ; multiple endpoints ; copula

Abstract

Modeling and simulation of correlated random variables is important for evaluating operating characteristics of various designs that involve multiple endpoints, some of which may be discrete (e.g., number of events) and some continuous (e.g., change of a continuous score). There exist efficient algorithms to address the problem of generating multivariate distributions with given marginals and given correlation structure, in particular NORTA (NORmal To Anything, Cario and Nelson (1997)). For numerical implementation of such algorithms, it is important to know the extreme values of pairwise correlations, which may be less than 1 in absolute value. We provide closed-form expressions for several classes of multivariate distributions that involve both discrete and continuous endpoints and illustrate the performance of algorithms via several examples.