2017/10/31 by Kyu Ha Lee, Brent A. Coull, Lee, Kyu Ha +7 · 1 citation
Mathematics · Computer Science · #Statistical Methods and Bayesian Inference #Bayesian Methods and Mixture Models #Statistical Methods in Clinical Trials
paper · pdf · doi:10.48550/arxiv.1711.00157
Microorganisms play critical roles in human health and disease. It is well\nknown that microbes live in diverse communities in which they interact\nsynergistically or antagonistically. Thus for estimating microbial associations\nwith clinical covariates, multivariate statistical models are preferred.\nMultivariate models allow one to estimate and exploit complex interdependencies\namong multiple taxa, yielding more powerful tests of exposure or treatment\neffects than application of taxon-specific univariate analyses. In addition,\nthe analysis of microbial count data requires special attention because data\ncommonly exhibit zero inflation. To meet these needs, we developed a Bayesian\nvariable selection model for multivariate count data with excess zeros that\nincorporates information on the covariance structure of the outcomes (counts\nfor multiple taxa), while estimating associations with the mean levels of these\noutcomes. Although there has been a great deal of effort in zero-inflated\nmodels for longitudinal data, little attention has been given to\nhigh-dimensional multivariate zero-inflated data modeled via a general\ncorrelation structure. Through simulation, we compared performance of the\nproposed method to that of existing univariate approaches, for both the binary\nand count parts of the model. When outcomes were correlated the proposed\nvariable selection method maintained type I error while boosting the ability to\nidentify true associations in the binary component of the model. For the count\npart of the model, in some scenarios the the univariate method had higher power\nthan the multivariate approach. This higher power was at a cost of a highly\ninflated false discovery rate not observed with the proposed multivariate\nmethod. We applied the approach to oral microbiome data from the Pediatric\nHIV/AIDS Cohort Oral Health Study and identified five species (of 44)\nassociated with HIV infection.\n