Slipping Away: Jo Aubin Has Alzheimer's. He's 38

Re:Slipping Away: Jo Aubin Has Alzheimer's. He's 38 by kpmah on Monday September 07, 2015 @04:30AM (#10180557)

The incidence of dementia for those aged 30-64 is about 54 per 100000. If you have a test that is 90% accurate and someone tests positive, according to Bayes you have 0.5% chance of actually having it.

Testing like that seems pointless to me.

Working: (0.9 * 0.00054) / (0.9 * 0.00054 + 0.1 * 0.99946)

[ Parent | on HN ]

Re:Slipping Away: Jo Aubin Has Alzheimer's. He's 38 by Phemist on Monday September 07, 2015 @07:56AM (#10181075)

Except that you are not picked from the whole population at large, but from a subset of it. Namely the subset that (weakly) believes they have dementia. This changes the incidence rate (or likelihood function) probably by at least an order of magnitude.

[ Parent | on HN ]

Re:Slipping Away: Jo Aubin Has Alzheimer's. He's 38 by kpmah on Monday September 07, 2015 @10:23AM (#10181515)

You're right. You have to update the probability with all the knowledge you have. Although I would say self diagnosis has a very, very high false positive rate, and if you know you're a hypochondriac that would make the probability drop, plus we don't even have a test for this that is 90% accurate. I'm not sure the difference is an order of magnitude :)

I think my point stands that it doesn't make sense to test for extremely rare diseases because of a self diagnosis.

[ Parent | on HN ]