Friday, January 7, 2011

So many studies, so few experiments

The African circumcision trials have yielded new data and analysis; Remember those? Foreign researchers attracted African men who wanted to be circumcised and get free health care, counselling, and follow-ups. Circumcision was immediate or in a couple years, at random. Great deal if you want the surgery. Thousands did.

The researchers collected so much data they've been able to mine it over the course of half a decade--and there's no indication they're done--for subsets with which to publish a paper, make news, and give press interviews. Let's face it--The researches want circumcision to work as much as the participants want a free circumcision--or more. They designed the tests and questions, collected the results, and are probing the data for evidence of efficacy.

Sound familiar?

Circumcision helps stop wart virus, study finds

Will the datamining ever end?

Bad methodology is supposed to be moderated by the understanding that you don't rely too heavily on any one experiment. But what if one experiment can be portrayed as dozens of studies, and publicized over a period of many years, giving the impression of many different experiements pointing in the same direction?

If researcher bias and experiment design can shave just a few percent off a measure in an intervention group, then an impressive looking relative reduction can be announced. Why waste that on just one event?

The researchers who so strongly hoped circumcision could be an answer to anything or everything were fortunate to lack the resources to publish all the results of their Ugandan genital cutting experiment expeditiously. That might have influence just one news cycle. Due to slowly dripping out their data, a poorly informed public fed by uncurious journalists will believe they are distinct experiments with independent findings when in fact there was just one experiment.

And all the ways they looked at the data and found the opposite effect, or no effect? They are under no obligation to publish them. Only confirmations of their grand hypothesis, that millions of years of evolution of male genitalia are a dangerous mistake requiring massively funded surgical campaigns, make it out of the desk drawer. Or, perhaps the data which they don't care for will be published, but as a lower priority. They'll get to at in a few decades, after they've circumcised every man in Africa and every newborn child in the United States.