Methodology and Causality in Religion and Health Research

The talk will survey some of the methodological challenges in assessing causality in religion and health research. While the field has seen increasing use of better longitudinal designs in assessing the relationship between religion and health, for certain questions considerable methodological challenges remain. For example, while it is the case that religious participation is longitudinally associated with less depression, it is also the case that those who become depressed are more likely to cease attending religious services. Such feedback, with causality in both directions, renders cross-section data useless, and requires causal modeling techniques, such as marginal structural models, to adequately assess the extent of the effects in both directions. Analyses using marginal structural models with the Nurses' Health Study data will be presented in studying this feedback between religious service attendance and depression. Another set of methodological challenges arises when interest lies in studying mechanisms governing the relationship between religious participation and health outcomes. The relevance of the literature on causal mediation analysis to religion and health research will be discussed generally, and analyses will be presented examining the mechanisms for the religion-mortality relationship in the Nurses' Health Study data.