you can peak every time if you do it Bayes — 100% no p-hacking guaranteed 

-
-
Replying to @o_guest @timtriche and
You can peek, but there is a multiple comparisons problem, because of high probability data will be misleading.
2 replies 0 retweets 2 likes -
-
Replying to @o_guest @timtriche and
If you use a BF > e.g., 3 to conclude there is an effect, you will have a multiple comparison issue.
1 reply 0 retweets 0 likes -
Replying to @lakens @timtriche and
Do you even have to use BFs? Aren't they just something p value fanatics crave regardless of usefulness?
1 reply 0 retweets 2 likes -
Replying to @o_guest @timtriche and
Which Bayesian approach did you have in mind? And why would multiple comparisons not be a problem?
1 reply 0 retweets 0 likes -
Replying to @lakens @timtriche and
I'm hoping you can explain your thoughts more to me — here's someth on mult comps, I'm sure you're familiar w/ it: http://www.stat.columbia.edu/~gelman/research/published/multiple2f.pdf …
2 replies 0 retweets 1 like -
Replying to @o_guest @timtriche and
Yes, see http://www.uni-bielefeld.de/lili/personen/jruiter/downloads/statisticsworkshop/Schoenbrodt_Sequential_BF.pdf … by
@nicebread303 . Gelman's solution is close your eye's, go 'lalala' and pretend type 1 errors don't exist1 reply 0 retweets 3 likes -
Hmm, I'm a bit disappointed — got anything that's actually a coherent piece that doesn't require me to guess what was said along w/ slides?
-
-
Could you point me to where
@nicebread303 says multiple comparisons are a problem? Slides seem to imply it's fine — no?1 reply 0 retweets 0 likes -
See https://osf.io/w3s3s . Multiple peaking increases rate of misleading evidence, but only towards an upper limit (not 100% as with p)
0 replies 1 retweet 4 likes
End of conversation
New conversation -
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.