Every Monday, Ph.D. students in the public health program at Brown gather to eat pizza, rearrange some unappetizing caesar salad around on our plates, and discuss a recent manuscript in different fields in an entertaining hour known as Journal Club.
Today’s article of choice was written in American Journal of Health Promotion, linked here, which promoted the idea that short bouts of moderate to vigorous exercise each day were successful in reductions of BMI. The article was titled “Moderate to Vigorous Physical Activity and Weight Outcomes: Does Every Minute Count?”
Methods, covariates, and study population limitations aside, what struck me as uncomfortable was how, despite the author’s self-admittance that this manuscript did not show causes and effects, the journal still placesets the following highlighted box.
In other words, “we can’t claim causation with our exercise exposure, but we urge you to change your lifestyle anyways.” Is that really worth a box? In my opinion, there’s too much uncertainty to simplify most statistical analyses into boxes, much less ones with the assumptions and faults of linear regressions.
While I’m at it, it also bugged me that the “significant” health effect which the article finds is a BMI reduction of 0.04, which, from my brief calculation, is less than a quarter of a pound for a normal male’s body weight, and quite possibly the result of the authors’ binning together several years of data for a large sample size. Okay, rant over.
StatsbyLopez 
Reblogged this on Stats in the Wild.