Hacker News Reader

The article doesn't say anything about Pr(Dislikes Rewards), but rather claims to have measured P(Diminished Response to Rewards | Clinical Depression Diagnosis) > P(Diminished Response to Rewards | !Diagnosis).

It's an 84 person study, done using EEG (a very rough measure of electrical activity in the brain), claiming a result largely related neurotransmitter release (ie not the sort of thing EEGs are usually used to measure). You should take their claims with a huge helping of salt.

Yeah my focus wasn't on the wording, it was on the order of the two variables. Interesting, thanks.

The first one. They sorted the kids into Depressed and not-Depressed before measuring the reward response.

The second probability depends on the relative sizes of all four quadrants and your definition of "high". The truth of statement one does imply that it's higher, but how much higher is not discussed.

Gotcha. But then what we really want is the second one, right? Which implies we might actually not be able to judge depression based on this, right?

Nice statistical property: this is the same thing.

P(B|A)>P(B) <=> P(A|B)>P(A)