Hacker News new | past | comments | ask | show | jobs | submit login

But they're just not -- that's exactly my point. If you test a bunch of Harvard undergrads (the most common subject type in these studies) -- that's not a uniform sample. In any way whatsoever.

Furthermore, even if you get a high p-value ... your experimental design can introduce so much noise.

Even when you do find these high p-value correlations -- they don't really ever say much about "why" ... because they can't. So much of that is the study design.

Let's say I make a study on how well people navigate a maze under the influence of alcohol vs. not. You can probably get a high p-value that they do worse while under the influence of alcohol... but it reveals very little about why.

So much of that is dependent on the particular maze. In fact I bet I could design a maze specifically to prove whatever conclusion I wanted. That is the whole problem in psychology. There is very little focus given to actually characterizing cognition in any meaningful way.

I like the clinical psychologists because they attempt to do exactly that, even though they have less "scientific" findings.




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: