One day you will read in the news that eating eggs has been linked with cancer and you shouldn’t eat them. Then six months later you will hear that they are a better cure for what ails you than unicorn tears – which makes most people think that scientists have absolutely no idea what they are doing. Is salt worse than cyanide? Or is it a miracle cure? JUST TELL US THE TRUTH!
Well, science is hard. Like, really hard. Designing a study, controlling as many of the variables as you can think of, and ensuring the compliance of the study subjects is hard enough when you are dealing with rats. When you add humans into the mix it gets so hard that I’m surprised that anyone even bothers.
Also, what is even harder than regular science, is statistics. A lot of the studies that are quoted in the media, and even the ones that are published in prestigious journals like Nature and Science, can have glaring statistical errors, omissions, or things that don’t make sense. Many results that are “statistically significant” are in all practical terms, pretty much a waste of time. Conversely, many effects that don’t reach statistical significance might actually be of use. But all of this is really hard to know, even if you have been trained in statistics and study design. If you want to learn more of this kind of thing, Alan Aragon’s Research Review is a really good resource. Also, Statistics Done Wrong is a really good layman’s resource.
What generally happens is a study gets published, some intern getting paid exactly nothing and with the scientific knowledge of pond scum reads the abstract or a press release and says “hey, there’s a story here!” The journalist then reads the summary of the abstract, and writes an article on it. This is how “in elderly untrained subjects (n=17), increasing protein intake above deficiency levels via ingestion of whole eggs increases protein synthesis” turns into “SCIENCE SAYS: EAT EGGS AND PUT ON 10LB OF MUSCLE TODAY!”
Or someone took the data from a huge long term study of like 60,000 people over 20 years, and found out that red meat consumption is correlated with a statistically significant increase of heart disease risk. Well, it also turns out that per capita cheese consumption is almost perfectly correlated (r=0.9471) with the number of people who die by becoming tangled in their bedsheets1. If you dump enough different variables into a big bucket of data, some correlations are going to drop out the bottom. Doesn’t mean that they actually matter.
So what does this mean for you? Well, it means that you can probably trust that anything you read about nutrition or exercise in the mainstream media is complete crap. At best. If you do the exact opposite of what they tell you to, you probably won’t be in any trouble at all.
If someone is telling you that you need to do anything other than eat a diet consisting of mostly whole foods most of the time, and try to not sit on your ass all day, they are selling you something. Just watch what commercials after that segment to see what it is.
- http://www.tylervigen.com/spurious-correlations ↩