4 ms·
The core issue is that p-values are cheaper to get than replicating the study, but replicating the study is the only reliable way to see if it's true or not. S
by rossdavidh 9y ago
The core issue is that p-values are cheaper to get than replicating the study, but replicating the study is the only reliable way to see if it's true or not. Sometimes the expensive/time-consuming way, is the only good way.
- lisper 9y agoReplication by itself is not enough. You need pre-registration too. Otherwise you can p-hack the replications.
- rossdavidh 9y agoTrue, but the article was making a point about "even if we set aside p-hacking for a moment", so I was going with that. But I agree, pre-registration is important as well.
- o2348diuu 9y agoNot really disagreeing with you necessarily, but you can hack pre-registration too, in a way that becomes more tenable with smaller studies, which benefit less from pre-registration anyway. The tables are misleading too: of course p(h0=T|p<.05) is going to be about 0.50 if your power is 0.50 and p(h1) = 0.10. That is maybe the bigger problem than p-values per se. Also, he fails to compute the other cell, which is p(h0=T|p>.05), which ends up being about 0.95. The problem in that table is have no power to detect anything. In real life, you don't know p(h0)--that's the point of doing the study. So Bayesianism doesn't really help you. The real problem is lack of power.
- tomrod 9y agoPre-registration?
- lisper 9y agohttp://www.apa.org/science/about/psa/2015/08/pre-registration.aspx http://www.apa.org/science/about/psa/2015/08/pre-registratio...