Using Random Spline Coefficients to Estimate Non-Constant Factor Loadings

• Activity Number: 314
• Type: Contributed
• Date/Time: Tuesday, August 11, 2015 : 8:30 AM to 10:20 AM
• Sponsor: Section on Statistics and the Environment
• Author(s): Zhenzhen Zhang*
• Companies: University of Michigan
• Keywords: latent factor model ; random spline coefficients ; likelihood ratio test ; EM algorithm ; simulated likelihood

Abstract

Latent factor models are commonly used to summarize multivariate exposure data. However, the factor loadings in these models are typically set as constant, implying a constant correlation structure. Multiple group comparison can be used to assess the validity of this assumption. In this paper we consider latent factor models with non-constant factor loadings estimated using random spline coefficients. We test whether a factor loading is constant through likelihood ratio test of the variance of random spline coefficients. The model is applied on maternal lead exposure data from the ELEMENT study to test whether the correlation among different lead measurements change with maternal age.