Bayesian analysis..my brain hurts

[Brown and Boris]

Had this thrown at me today and don't understand a word. Perahps they were just trying to show how clever they are and how stupiod I am. Well they suceeded.

It was trying to calculate the probability that a test would falsely assign people to a group as likely to do something bad (like resume smoking)


"Bayesian analysis reveals that using test scores to predict future behaiour will yield false assignment
rates between 60% when the test failure rate (ie says they will start smoking again) is 70%, the base rate of behaviour (resuming smoking after giving up) in society is 30%, and the test hit rate is 90%)

( I have no seprated the nextt bit as I think it was a confusing add on bit.....


and 90% (when the test failure rate is 90%, the base rate ofthe behaviour is 10% and the test hit rate is 70%)."

How have they got the 60% from the 30% 70% and the 90%. Blah Blah.

Anybody calculate the probability where the % expecting to start smoking again is only 1%, 5% or 25%.


They also used words like athoretic, stastist interventionist and Potemkin village cloaks which I thought was something people in Norfolk wore.

[RupertTheFridge]

http://en.wikipedia.org/wiki/Bayes_estimator


might help

[Brown and Boris]

RupertTheFridge said:

http://en.wikipedia.org/wiki/Bayes_estimator


might help

I had found that, but thanks anyway.

It seesm to be some simple probability equation, that if say 20% of people who give up are known to start again, and that the test in fact identifies 50% of people who it says will restrat, that suggests that 60% of that 50% will be false?

Based on that where we know the restart rate is 20% and our test identifies about 20% bizarrely that suggests we have a 100% hit rate but that isn't true because some of those people wil not retart and some we have no identified will strat, so it seems pretty blunt.