Sparse Logistic Normal Regression for Modeling Over-Dispersed Count Data with an Application to Microbiome Data Analysis

• Activity Number: 469
• Type: Contributed
• Date/Time: Wednesday, August 1, 2012 : 8:30 AM to 10:20 AM
• Sponsor: Biometrics Section
• Author(s): FAN XIA*+ and Hongzhe Li
• Companies: The University of Hong Kong and University of Pennsylvania
• Keywords: Compositional data ; Logistic normal regression ; Group variable selection ; MCEM ; Human gut microbiome

Abstract

The human body contains trillions of microbial cells which have an influence on human health and predisposition to complex diseases. One important problem of microbiome data analysis is to identify the environmental covariates associated with different taxa. However, the taxa count data is often over-dispersed and therefore we propose to use a hierarchical statistical model that combines the Aitchison's logistic normal (LN) distribution for the latent compositions and conditional multinomial distribution for the observed counts. To deal with the high dimensionality of the covariates, we apply a group l1 penalized LN regression for the selection of relevant covariates. This model can naturally accout for sampling variabilities and zero observations. We present a Monte Carlo expectation-maximization (MCEM) algorithm to estimate the model parameters. Our simulation results show that the pro