In the comments section of a recent post on religiosity, a reader linked to a study which found that young adults who attended church weekly were 50 percent more likely to be obese when measured again 18 years later.
GSS interviewers assessed the weight of respondents (below average = 1, average = 2, above average = 3, considerably above average = 4) and asked about frequency of church attendance. Here are standardized OLS regression coefficients:
DV = respondent's weight (sample size = 2,256)
Church attendence .02
* statistically significant
First of all, the bivariate relationship between church attendance and weight is zero. It remains zero with all the controls. Older and less educated people and blacks tend to be heavier. Gender, living in the South, and church attendance are unrelated to weight.
I would still side with the AHA study since weight is measured more precisely in that study, and the data are longitudinal.