Chunk #17 — Methods — Discovery Process of Environmental Factors Contributing to ERS for Single Phenotype — 3. Conditional analysis via multi-pollutant models
Motivated by the discovery strategy of additional genetic loci via conditioning on the loci identified through marginal association in GWAS [37], we further explored the possibility of identifying additional pollutants not selected in the previous two-stage analysis, in the presence of the previously selected ones in a multivariate model. Specifically, for subject i, let Ei+ denote a pollutant not belonging to Es. The conditional model is given by.(3)where . We repeated the two-stage analysis with this conditional model for each pollutant not belonging to Es. We calculated the same Wald test statistics and compared them to the same thresholds to select additional exposures for ERS. We denoted the set of pollutants selected in this step as Ec, denoting pollutants identified based on conditional analysis.
