Follow the science! Or not….
No End to Parallel Universes in Nutrition
How about a multiverse where nutritionists always get exactly the answer they want from their studies?
As Alpha Waymond might put it, sure.
This scenario is in fact our reality, according to a recent analysis published in The Journal of Clinical Epidemiology. Nutrition science, and particularly nutritional epidemiology, has a multiverse problem, the paper asserts, and it’s yet another reason to be very skeptical of this research.
/snip
At the heart of the multiverse problem in nutrition is a simple concept known as “analytical flexibility.” No consensus exists in nutrition science about how best to analyze data from the numerous epidemiological surveys that have appeared in the last 60 years to try to make sense of the diet-lifestyle-disease relationship. The epidemiologists have never done the necessary hard work to establish which of any number of strategies and techniques is reliably correct.
Any analytical strategy, as Zeraatkar explained in an interview, can involve hundreds, if not thousands of decisions, large and small, about how best to conduct an analysis.2 Hence, the uncertainty in the diet-health associations that these studies produce is not just about whether we can trust, for instance, that the self-reported dietary data accurately reflects what people really ate, but all the particular decision points made by the researchers doing these analyses. A different analytic technique or different decisions about how best to handle the data can produce different answers – a different multiverse of answers, in effect. Use one set of criteria, and the diet or lifestyle factor of interest will be associated with an increased risk of a particular disease; use a different set of criteria, and the reverse appears true. How do we know which is right? Which multiverse is our reality, the one we’re living in?
/snip
One obvious implication was that researchers looking for publishable results could easily make the decisions – consciously or subconsciously – to get the results they wanted, whether or not that was the correct result.3
Leave a Reply
You must be logged in to post a comment.