Corrections to the blogosphere, the consensus, and the world

Friday, July 27, 2007

Lovaas's best correlation

From Autism Diva:
"The correlation between miles between the Lovaas Insitute and the top five autism-serving RCs in driving miles and how high their percentage of clients served in the autism category is represented by a Pearson's r of .89.

Wow and begorrah. Aut Div, that's far and away the highest correlation I've ever seen for anything in the autism area, bar none. I'm really impressed.

If I could point you at another aspect of the picture, my fundamental beef with Lovaas (leaving the ethics out of it for the moment) is that his proof of the efficacy of ABA seems to miss the point completely (and this carries with it the corollary that nearly all the attacks on the efficacy of ABA also miss the point completely) by focusing on the one or two or three or four or five, whatever, published studies. All very well, and Michelle Dawson and her mob have done a marvellous hatchet job on it all, but I would have thought largely beside the point.
The hole in Lovaas is that the Lovaas centres must have data on anywhere between two and five thousand cases - diagnosis, sex, treatments, outcomes - because it'll be in their management database (or their filing cabinets, considering we're looking back over thirty years). They must have figures on length of treatment, intensity, school placement, and cost, because all of those things are under their control and must be recorded. And they haven't released any of it.

Coming from health promotion, my bias is very heavily towards preferring epidemiological data to clinical studies. Coming from a legal background my bias is to believe that if Lovaas has clinical studies and administrative data and releases one but not the other it's because one is more favourable to his position than the other. I'm not inclined to give Lovaas any credence until he tells me what he knows that I don't.

No comments:

Blog Archive

Search This Blog

Followers

Follow by Email

There was an error in this gadget

Total Pageviews