<$BlogRSDUrl$>

Tuesday, August 14, 2007

*
*
Why the Internet Autism Test is probably not reliable

A test of one's Autism Quotient is readily available on the Internet. Its most famous appearance may be in an article in Wired magazine, which included a copy of the test. Many people have taken the test and seem to take a perverse pleasure in scoring high, thereby identifying themselves as somewhat autistic and therefore as true "geeks".

Reference to a scholarly paper by those who developed this test makes me suspect that the self-diagnostic approach to the test is flawed. Checking the fine print, we find this:

" ... 80% of adults with AS or high functioning autism scored above a critical minimum of 32, whereas only 2% of control adults did. "

What this is saying, as far as I can see, is that people known to have autism generally scored high on a test designed to test for autistic traits. OK. But it is then noted that 2% of control adults scored high as well. What does this imply? That 1 in 50 people in the ordinary population will score as "autistic". Autism is generally considered to be a rare condition. But 1 in 50 people is a very high number of people. Clearly this test will produce a lot of false positives, with large numbers of people with no clinical condition scoring into the abnormal range associated with autism.

One of the signs of autism is social withdrawal. However to be shy and socially withdrawn - to be scholarly in other words - is not to be autistic. I feel that people are testing themselves for a symptom of a disease, and then deciding that they have the disease. It's a form of hypochondria.

To use an analogy, perhaps 2% of the male population is over six feet two inches. Men with excess growth hormone will mostly be over six feet two inches. But that does not mean that everybody over six feet two inches has a hormonal problem.

Julian

|

Friday, August 10, 2007

*
*
The New Inventors

"The New Inventors" is a program on Australian television. I saw an episode tonight as part of an otherwise excellent line-up on the Discovery Science Channel. Something about "The New Inventors" has been increasingly bothering me, and this latest offering crystallised the problem.

There were three inventions, one of which was a "cruise control" for a four-wheeled agricultural motorcycle, of a kind used to spray fruit trees and the like. There was nothing much wrong with this, except that it hardly seemed novel. The other two inventions were a pole-like device that purportedly measures features of snow that might be correlated with the risk of an avalanche; and a device to be worn outside one's clothing on a sunny day to warn when one has received too much ultraviolet light. My problem with both these latter inventions was that none of the panelists asked any questions about how the devices had been calibrated or tested. The lady engineer on the panel asked if the sensor that was to be driven into snow to test for, presumably, compaction and layering had to be put in close to the vertical. But there was no attempt to determine that the device was actually capable of its purported function - predicting avalanches. How many false negatives and false positives were likely? An important question I would have thought.

Turning to the ultraviolet light warning device, again there were no questions as to whether it had been shown to reliably integrate ultraviolet light exposure. How much exposure had been determined to be enough? How did the memory aspect of the device, which supposedly remembered your exposure on previous days, actually work, and what was its theoretical basis? Was it intended to determine how tanned you had become and therefore that you might be able to tolerate more ultraviolet light? If so, what if you were wearing a shirt one day, and no shirt the next? How would the device "know" that?

"The New Inventors" makes a bit of a habit of this half-baked approach. A flashy invention wins, but later one starts to have one's doubts.

Julian

|

This page is powered by Blogger. Isn't yours?