Regression Analysis of Dependent Binary Data for Estimating Disease Etiology from Case-Control Studies

• DOI: 10.48550/arxiv.1906.08436
• Publication Date: 2019-01-01
• Authors: Zhenke Wu; Irena Chen
• Keywords: Etiology

In large-scale disease etiology studies, epidemiologists often need to use multiple binary measures of unobserved causes that are not perfectly sensitive or specific estimate cause-specific case fractions, referred as "population etiologic fractions" (PEFs). Despite recent methodological advances, the scientific incorporating control data effect explanatory variables upon PEFs, however, remains unmet. this paper, we build on and extend nested partially-latent class model (npLCMs, Wu et al., 2017) a general framework for regression analysis in case-control studies. Data from controls provide requisite information about measurement specificities covariations, which is used correctly assign probabilities each given her measurements. We distribution controls' diagnostic covariates via separate priori encourage simpler conditional dependence structures. Markov chain Monte Carlo posterior inference PEF functions, cases' latent classes overall PEFs policy interest. illustrate with simulations show less biased estimation more valid than an npLCM omitting covariates. A childhood pneumonia study site reveals season, age, severity HIV status.