Addressing the causal ordering of religion and health, researchers are beginning to use more rigorous statistical techniques such as structural equation modeling and simultaneous equations. Using hierarchal linear modeling techniques, investigators are beginning to estimate the differential effects of individual religiosity and community religiosity. An example of community religiosity includes the number of nearby churches in individuals neighborhoods of residence.
Although these statistical techniques can establish more precise estimates of the direct and indirect effects of religion and spirituality on health and distinguish between the individual versus community effects of religion, these approaches do not eliminate selection bias. Selection issues arise, for example, if some participants go to church because of an underlying motivation to engage in healthy behaviors or they are experiencing a severe or terminal illness. If these health-related motivations to participate in church activities are not measured and are not included in statistical models, the positive effect of religion on health may be overestimated. Notably, one study reduces the scope of estimation error by controlling for several measures of motivation to participate in religious organizations (Franzini, Ribble, & Wingfield, 2005).