I have spent many years developing statistical methods to understand how to measure dietary intakes in a population and how to relate such measures to mortality and chronic diseases. In animal experiments, clinical trials of different dietary patterns have been performed by my colleagues at Texas A&M and show, in stunning fashion, that for example a fish oil enhanced diet is protective against colon cancer, DNA damage, deleterious gene expression, etc., when compared to a corn oil enhanced diet (think potato chips).

In humans, the statistical questions are much more difficult, because it is impossible, in current practice, to measure an individual’s long-term average dietary intakes across multiple foods and nutrients. This impossibility has many facets, which I will briefly review. This statistical issue, along with the media focus on dietary bullets (kale anyone?), has resulted in massive confusion, and sometimes silly conclusions. For example, what % of U.S. children eat an alarmingly poor diet: a simplistic analysis says 30%, but a more nuanced one that accounts for the measurement properties of dietary instruments says 8%, a massive difference. My aim in part is to show that focusing on dietary patterns, instead of magic bullets, leads to far more robust statistical conclusions than focusing on one potential bullet at a time.