Simultaneous Joint and Marginal Models Approach for Testing Multivariate Binomial Data

• Activity Number: 487
• Type: Contributed
• Date/Time: Wednesday, August 7, 2013 : 8:30 AM to 10:20 AM
• Sponsor: Biopharmaceutical Section
• Author(s): Shuling Liu and Kerry Go and Manoj Thakur*+
• Companies: Emory University and Sanofi Pasteur and Sanofi Pasteur
• Keywords: multivariate binomial ; simultaneous models ; correlation structure ; solicited reactions ; generalized estimation equation

Abstract

We consider simultaneous joint and marginal regression analysis of the multivariate binomial data. Our method is motivated by the comparison of solicited reactions in vaccine clinical trials in which a vector of binary outcomes is collected for treatment group and control group. Since those outcomes are correlated, it is natural to view the data as multivariate binomial responses with certain correlation structures. This type of data involves several challenging statistical features. For example, comparability between groups for "any reaction" does not necessarily imply comparability between groups for each particular reaction. To address these issues in a multivariate manner, we propose a simultaneous joint and marginal model framework for joint probabilities and marginal proportions where dependence among the reactions for the same subject is modeled via random effects. An EM algorithm to derive approximate maximum likelihood estimators has also been developed. Alternatively, the model can be fitted through the generalized estimation equation. Simulation studies are conducted to demonstrate the performance of our proposed approaches, and our method was applied to vaccine data.