Here's the study. It was looking at a 393-person subset of a larger study, which as I've said is really, really small for this kind of observational trial https://jamanetwork.com/journals/jamanetworkopen/fullarticle/2719576 …
-
Show this thread
-
Seriously, I doubt whether this was sufficiently well-powered to establish a statistical difference once you've split it up into so many categories
1 reply 0 retweets 1 likeShow this thread -
You can kind of see this in that they split the drinkers into 4 groups: never, abstainers, 1-7 drinks per week, 7+ drinks per week These are ~really weird~pic.twitter.com/d0yF2oA7Z5
1 reply 0 retweets 0 likesShow this thread -
Usually with drinking, you'd divide it into 1 per day (i.e. 7 per week), 2 per day, 3 per day etc This significantly limits the applicability of the findings straight up, because the AVERAGE alcohol consumption in many countries is more than double this 'safe' amount
1 reply 0 retweets 0 likesShow this thread -
And the tiny sample size means that their results are only ~barely~ significant, because the difference between moderate and other drinkers is always very small anywaypic.twitter.com/a7eMAqTIeP
2 replies 0 retweets 0 likesShow this thread -
Replying to @GidMK
I agree with your overall points here, but I'm not sure about the fixation on decision theoretics. Why would statistical significance or power matter here (in a study that's observational and has no random mechanisms)?
1 reply 0 retweets 1 like -
I'd say this is one of those scenarios where there is little justification for it (unlike experiments that can be replicated and constantly involve some sort of randomization, or surveys that involve random sampling etc.)
2 replies 0 retweets 1 like -
This seems to me like a scenario where interval estimation would shine as would interpreting any observed Ps in a neoFisherian way (as continuous measures) or using Bayesian methods like the ROPE (region of practical equivalence procedure)
2 replies 0 retweets 1 like -
I can say that JAMA is probably one of the worst journals with zero interest in proper statistical inference
1 reply 0 retweets 1 like -
are there any good online stats classes that either of you two would recommend. i had a good stats class in undergrad but forgot most of it and stats here is taught plug-and-chug and not critically.
4 replies 0 retweets 1 like
I've learned more about stats from Twitter and Googling my questions than from any of my uni courses tbh
-
-
Replying to @GidMK @kevinnbass
Well, I'm definitely going to have to plug my blog which is pretty much all stats and history of stats and folks like
@Lester_Domes have tremendously helped correct errors in it and his works have led to several blog posts too https://www.lesslikely.com/1 reply 1 retweet 4 likes -
Yes also read Zad's blog it's good
1 reply 1 retweet 1 like - Show replies
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.