Cause and Event: Supporting Causal Claims through Logistic Models

Efforts to identify and support credible causal claims have received intense interest in the research community, particularly over the past few decades. In this paper, we focus on the use of statistical procedures designed to support causal claims for a treatment or intervention when the response variable of interest is dichotomous. We identify seven key features of logistic regression studies that should play a critical role in estimating a causal effect and discuss their implications for causal inference. These include elaboration of research design, clarification of link function, model specification, challenges and limitations of sample size, interpretation of treatment effect through odds ratios, statistical tests and examination of model fit, and the potential for multilevel logistic models in pursuit of causal claims. Our recommendations are intended to guide researchers in the critical evaluation of logistic regression models for analyses culminating in causal claims and to promote stronger design and modeling strategies for reliable causal inference.