Finding the Causes of Disease
I admit it, I’m a bit of a stats geek. I have a graduate minor in statistics, and I enjoy reading well-done analyses of complex datasets.
Last week, Significance Magazine published a well written teardown of a study that claimed to link GMOs to various diseases. Spoiler alert: the original paper is a wreck of badly done analysis.
I’ll leave the full discussion to Mr. Johnson, as I can’t do a better job. However here’s a quick summary for the short on time:
- Confirmation bias occurs when someone tends to look for data that supports their underlying beliefs; the study authors have clear and documented bias against GMOs
- Correlation is not causation; a correlation says two things tend to increase or decrease at the same time, but says nothing about whether one causes the other, or there’s a common root cause, or it’s just coincidence. In this study, the authors rely on dubious correlations to support their confirmation bias.
- Graphs are a fantastic tool, and can so easily manipulate and mislead. The authors commit several offenses here. I highly recommend an aged but delightful (and quick) book called How to Lie With Statistics (Goodreads).
Demonstrating a dietary or environmental root cause of a disease can be challenging. Many progress very slowly, and it’s unethical to perform a truly controlled study on humans. Numerous confounding factors exist, meaning that study populations must be massive. Changing definitions and increased vigilance can artificially inflate the reported incidence of the disease.
It’s possible that there are links between GMOs and some diseases, but it has yet to be demonstrated with any reliable, unbiased study. If you come across a paper, website, or other source that does claim a link (or even one that definitively claims no link), please take a long look and remember that robust data analysis requires some training and experience in order to avoid incorrect conclusions. In other words, “extraordinary claims require extraordinary evidence.”