Thanks! This nicely sums up an issue with these food-frequency questionaires:

> admitting to eating ice cream might correlate with metabolic health

That is, ice cream might well be bad for you but if you experience the bad effects, you might stop admitting that you eat ice cream and thus add an opposite data point in the study.