Medicine: TV-Autism Smackdown/CounterSmackdown

So Time GOES OFF on the TV-autism study from last week, making a handful of excellent points mostly about the way that the media (statisticians and research scientists) make improper derivations from correlational studies. It is notable to say that the study hasn't actually been published yet, but these quotes are from press releases and the such:

"Approximately 17% of the growth in autism in California and Pennsylvania during the 1970s and 1980s was due to the growth of cable television," and "just under 40% of autism diagnoses in the three states studied is the result of television watching due to precipitation."

Result of? Due to? How can these researchers suggest causality when no actual TV watching was ever measured? "The standard interpretation of this type of analysis is that this is cause and effect," Waldman insists, adding that the 67-page study has been read by "half a dozen topnotch health economists."

Could there be something to this strange piece of statistical derring-do? It's not impossible, but it would take a lot more research to tease out its true significance. Meanwhile, it's hard to say just what these correlations measure. "You have to be very definitive about what you are looking at," says Vanderbilt University geneticist Pat Levitt. "How do you know, for instance, that it's not mold or mildew in the counties that have a lot of rain?" How do you know, for that matter, that as counties get more cable access, they don't also get more pediatricians scanning for autism?

Sure, and Wallis does make some good argument that going Freakonomic on health care statistics leads us all to use language that we can't use. We can't say that TV causes autism, but we can say they are very associated. In the case of this correlation, it seems most likely that a confounder affecting both variables probably gives us the R-squared, the "caused-by" variable, that we're seeing. Or maybe it's TV. Who knows.

But we have to give props to Gregg Easterbrook for SLAMMING Wallis and Time for what he thinks is absurd scientific idealism. Short enough to post in full:

This article is awfully casual about accusing others of "irresponsibility." It is common for ideas in research to begin with incomplete statistical observations that inspire National Science Foundation- or Institutes of Medicine-sanctioned studies to prove or disprove the underlying larger claim. The sort of research TIME takes two Cornell University professors to task for not having already conducted at their own expense would require millions of dollars for a very complex multi-year home-monitoring study with hundreds of families in the study group allowing two-way recording devices throughout their homes and hundreds more in the control group.

A correlation found during initial research is exactly the sort of event that triggers funding for such major studies. TIME declaring that a statistician who finds a clue should not publish unless he can offer definitive proof is like saying an astronomer who discovers a star should not reveal its location unless he can prove the origin of the universe. And suppose this theory of autism turns out to be true. Should those with suspicions remain silent, offering no caution to parents of young children?

At any rate, since TIME sees fit to accuse others of irresponsibility, it would have been nice if TIME's article had disclosed that its corporate parent has a financial interest in denouncing this research. TIME is owned by the same company that owns Time Warner Cable, a leading cable television carrier, and owns Cartoon Network, which is marketed to young children.

Ouch. I'm bleeding after reading that.

So let's blame who we should blame: the folks who publish "TV causes autism" crap all over their headlines to drive traffic to their sites, to sell newspapers, and even, oddly enough, boost ratings. The Cornell study might need an editor to throw in some subjunctive verbs. But this army of media douchebags who couldn't properly interpret a medical study if their Blackberrys depended on it need some good old fashioned Cuckoo's Nest-style ECT (before the widespread use of neuromuscular blockers that has made the procedure perfectly uneventful to watch).