Sunday, June 3, 2012

Cherry-Picking Statistics

Dr. Eenfeldt shared this image on his blog yesterday in a post called The Sugar Empire Strikes Back. It's a comparison between per capita soda consumption and the adult diabetes rate. Most low carbers see this as confirmation of their argument that sugar is harmful.

Eenfeldt first referenced it in a post called Soda and Diabetes – a Coincidence? and says that it came from this lecture by Dr. Robert Lustig.

I did a little digging around and it looks like Lustig pulled these maps from the the USDA Food Environment Atlas. It's actually a pretty nifty site and it's worth checking out.

Well, I found another map that also shows a pretty strong correlation with diabetes:

Hmmm... I wonder what it could be... Proximity to fast food restaurants? Food deserts? Snack consumption?

No. It's Meat & Poultry. Oops! Maybe we shouldn't be so quick to start throwing these epidemiological statistics behind our arguments. A correlation is not proof of causation and shouldn't be used to argue the veracity of a claim. While I agree with Lustig that soda is very unhealthy, I think it would serve his case better by sticking to actual evidence rather than hype.

Then again, maybe we have the causality all mixed up and it's really Meat & Poultry consumption that causes consumption of soda. Gotta wash down those burgers and fried chicken with something.

Apparently being a Californian or a Texan has a protective effect on that causal chain, though. ;)

The thing we should be concerned about is that if Lustig succeeds in using such tactics in convincing politicians that laws should be put in place to limit sugar consumption, it's only a matter of time before meat consumption succumbs to the same fate. Both sides of the debate can play this epidemiology game and I wouldn't be surprised if a map detailing butter consumption looked similar, too.


  1. As much as I appreciate your high school stats "Correlation does not imply causation" quip, I believe you are missing the point.Your second map does not have a correlation nearly as strong as the one comparing diabetes and soda consumption. California, a state that apparently has a very low concentration of diabetes, has a strong concentration of meat and poultry consumption. Also Mississippi and Louisiana, states that have so heavy a concentration of diabetes they are hard to see, are relatively meat and poultry free. There are many more comparisons that don't match up. I appreciate your concern for misleading statistics, but this meat and poultry comparison map is not a valid argument. Arguing against soda as being a major cause of obesity and diabetes will put you on the wrong side of this health battle. Also as long as I'm here, meat and poultry, as most Americans eat it, IS unhealthy. We eat far too much, and fast food, as the number one source of meat and poultry in our diets, has contributed just as much as soda to diabetes. This alone justifies any correlations between meat, poultry, and diabetes. I disagree that there is one, but if there is, it certainly makes sense.

  2. While I agree that the meat consumption map doesn't appear to show as strong of a correlation, it certainly is there. I could just easily start nit-picking regions where soda consumption and diabetes don't match up as nicely, either.

    Where did I suggest that soda isn't a major cause of obesity and diabetes? I even said that "I agree with Lustig that soda is very unhealthy". I'm sorry if I failed to make my point clear in this post because you appear to be missing it. I'm not arguing that soda isn't a cause of diabetes. I'm also not arguing that meat is. I'm pointing out the fact the using such correlations as proof is bad science.

    In my opinion, using bad science to support a good cause is a bad habit because we lose credibility every time we do so. That's all I'm trying to say. Even if your claims are correct and most of what you say to support your position is accurate, when you start use crap science, people stop listening.