Pure Pedantry : A Better Critique of DeSoto and Hitlan

Reality is always more complicated than you think.

Profile

Jake Young is a MD/PhD student at Mount Sinai School of Medicine focusing in Neuroscience. He is due to graduate in 2032. He received a BS and a MS in Biological Sciences from Stanford University -- where he spent most of his time drinking heavily and building vegetable catapults instead of learning information that would now be eminently useful. When he is not failing terrifically to perform his sworn duties, he enjoys watching bad movies, ethnic food, and running.

Pure Pedantry is a blog about science -- social sciences and otherwise -- as well as academic and scientific culture. No one can live on science alone, so I also like to dwell on pop culture, periodically explore the humanities, and indulge in other types of geeky goodness.

Jake is joined periodically by two wonderful guest bloggers: Kara Contreary and Kate Seip. See the About Page.

DISCLAIMERS: 1) Jake Young is not a licensed physician (yet). He is merely a medical student. The information published on this site is not intended for use in medical decision making. Please seek advice from a licensed, medical professional before making any health decisions. 2) The opinions expressed are my own or those of my co-bloggers. They do not represent the views of SEED magazine or the educational establishments we currently attend.

Search this blog

A Better Critique of DeSoto and Hitlan

Category: Autism • Statistics
Posted on: July 16, 2008 7:31 AM, by Jake Young

Not to reignite the whole issue with DeSoto and Hitlan (here and here), but Epi Wonk emailed me an excellent critique of the paper.

(The whole issue was about the relationship between mercury and autism. Ip, 2004 measured mercury levels in children with or without autism. They found no relationship between autism and mercury concentration. DeSoto and Hitlan corrected and reinterpreted their data and claimed that a correlation existed.)

Epi Wonk makes two very important points:

First, both Ip, 2004 and the revision in DeSoto and Hitlan, 2007 used a Student's t-test to analyze the data when the data does not appear to be normally distributed. Here is a key graph of the raw data (from Epi Wonk):

It breaks down the percentage of cases by mercury concentration. To quote Epi Wonk: "If these data are normally distributed, or anything close to normally distributed, than I'm Bernadine Healy." Thus, a t-test is inappropriate.

Second, using odds-ratio you can show that the apparent statistical significance is the result of the small percentage of very high mercury cases with autism. People pushing the autism-mercury link may jump on that, but I ask them: how do you explain the very large number of autistic children with very low levels of mercury?

Anyway, read the whole thing.

Find more posts in: Medicine & Health