If you haven’t read my Will Eating Red Meat Kill You? guest post on Mark’s Daily Apple from a few months ago (which I can’t hold against you, since I never posted the link!), I recommend hopping over there right now to skim the section on food frequency questionnaires. (If two long blog posts are too much Denise for one day, check out this classic piece by Chris Masterjohn on the unreliability of self-reported data instead.)
The bottom line—discussed more thoroughly in those aforementioned links—is that bad things happen when you ask people to report what they eat. That badness only amplifies when they’re reporting from memory, and becomes incrementally worse the further back in time you want them to recall. Most folks just aren’t that aware of what goes in their mouths, especially when those things went in their mouths six months ago. As a result, nearly everyone underestimates what they really consume, and some foods—usually the ones we think are bad for us—tend to be a major roulette-wheel spin in terms of accurate reporting. Health-conscious folk may be particularly likely to underreport their intake of “bad” foods out of sheer guilt (who wants those Saturday brownie binges emblazoned forever on paper?). And since not everyone misreports their food intake in a consistent way, no amount of statistical wand-waving can really make up for the inaccuracies in this type of data.