Statistics, Semantics and Parity
Posted by Miriam Gordon on March 19, 2012
I am by no means a statistician, but I’ve worked as a medical writer and editor for long enough to know that the results of clinical trials rely heavily on statistics for their interpretation. There have been countless books written on this matter, and here’s a gem of an article in the Atlantic that sums it up nicely. But since most of us are not statisticians, we must rely on the “experts” to interpret them for us. The problem is that they can be so easily misused and misleading.
If you recall, there was a famous study by Christakis and Fowler (NEJM 2007) on the spread of obesity through social networks that relied on statistical analysis. In 2008, a refutation of this article was published by Cohen-Cole and Fletcher in the Journal of Health Economics. In fact, I would bet you a pack of bubble gum that for every statistical analysis ever published, a refutation of that analysis is subsequently published. So what are the non-statisticians to do?
I say we all need to learn to take most studies with a grain of salt, and approach them skeptically. This is something that children should be learning in school – it is essential to the scientific process. If this concept were something that all of us learned in school, we wouldn’t be so quick to jump on the published results of any study even if they would prove our dearly held beliefs.
NB: In their blog “Retraction Watch,” Ivan Oransky and Adam Marcus have touched on the problems caused by faulty statistical analysis in climate change research. I hope to see more of this in the future. I attended a conference 2 days ago entitled “Bioethics Bootcamp,” sponsored by the Hastings Center, Science Writers in New York, and the National Association of Science Writers. Ivan, the Executive Director of Reuters Health, said that he routinely tells his reporters to “always carry a statistician in your back pocket.” Easier said than done, but I hope this will get easier in the future.