A Bayesian Confirmatory Factor Model for Familial Data with Multiple Outcomes

• Activity Number: 127
• Type: Contributed
• Date/Time: Monday, August 4, 2014 : 8:30 AM to 10:20 AM
• Sponsor: Biopharmaceutical Section
• Author(s): Qiaolin Chen*+ and Robert E. Weiss and Catherine A. Sugar
• Companies: Novartis and University of California, Los Angeles and University of California, Los Angeles
• Keywords: Bayesian family factormodel ; Confirmatory factor analysis ; Markov chain Monte Carlo ; Multiple outcomes ; Full information maximum likelihood ; Multi-trait multimethod matrix

Abstract

The UCLA Neurocognitive Family Study (NFS) collected multiple measurements on schizophrenia (SZ) patients and their relatives, as well as control subjects and their relatives. The relationship structure is complicated because not only observations on individuals from the same family are correlated, but the multiple outcome measures on the same individuals are also correlated. Traditional familial data analyses model outcomes separately and thus do not provide information about the interrelationships among them. I propose a Bayesian Family Factor Model (BFFM), which extends the classical confirmatory factor analysis (CFA) model using a combination of family-member factors and outcome factors. Simulation studies show that the traditional methods for fitting CFA models, such as full information maximum likelihood estimation using quasi-Newton optimization (FIML-QNO) often fail to fit the data due to non-convergence problems and Heywood cases caused by empirical under-identification. In contract, BFFM provides stable estimates. When both methods successfully fit the data, estimates from the BFFM have smaller variances and comparable mean squared errors.