Complementary Log Regression for Sufficient-Cause Modeling of Epidemiologic Data.

ABSTRACT: The logistic regression model is the workhorse of epidemiological data analysis. The model helps to clarify the relationship between multiple exposures and a binary outcome. Logistic regression analysis is readily implemented using existing statistical software, and this has contributed to it becoming a routine procedure for epidemiologists. In this paper, the authors focus on a causal model which has recently received much attention from the epidemiologic community, namely, the sufficient-component cause model (causal-pie model). The authors show that the sufficient-component cause model is associated with a particular 'link' function: the complementary log link. In a complementary log regression, the exponentiated coefficient of a main-effect term corresponds to an adjusted 'peril ratio'

SUBMITTER: Lin JH 

PROVIDER: S-EPMC5154187 | biostudies-literature | 2016 Dec

REPOSITORIES: biostudies-literature