-
-
Replying to @GidMK @Rootclaim
1. The mistake was in using 13/50 instead of 13/26 and in comparing probabilities rather than odds. 2. We verified it with the authors. 3. Please provide an example where a small sample fails despite a strong p-value. 4...
1 reply 0 retweets 1 like -
4. Where is the challenge in giving to 50 hospitalized patients a safe oral treatment every few days? Why would anyone drop out? 5. Search "multivariate logistic regression" in the paper. 6...
2 replies 0 retweets 1 like -
6. The point is that in a pandemic, we don't have the luxury of waiting for absolute certainty, we need to make the best decision for patients using the best information we have. 7. Please do not block people making an honest scientific discussion on matters of such importance.
1 reply 0 retweets 3 likes -
Replying to @saarwilf @Rootclaim
If you don't want people to block, I'd suggest refraining from tweeting the same tweets several times. I just automatically block that these days, it's very annoying
2 replies 0 retweets 0 likes -
Anyway, more broadly you haven't actually addressed anything I've said aside from the ICU admission point. If you've confirmed with the authors, I'd recommend that they request a revision from the journal specifying that this was the case because it is currently unclear
1 reply 0 retweets 2 likes -
The p-value point is a misunderstanding of the problem with small samples (it's not all about 'significance'). This is mirrored in the response to point 5 - this sample is FAR too small to run a multivariable logistic regression, all you'd get is noise if you adjusted
2 replies 0 retweets 2 likes -
Replying to @GidMK @Rootclaim
It's the same: It is indeed hard to reach significance with logistic regression on small samples, but if the effect is strong, it can work. In this case it is, reporting 95% CI of 0.003−0.25. If this study's conclusion is wrong, it's definitely not due to the small sample size.
1 reply 0 retweets 0 likes -
Replying to @saarwilf @Rootclaim
It's not hard to reach statistical significance at all with small samples and logistic regression. What's hard is to see a meaningful result once you add in numerous 'control' variables
1 reply 0 retweets 0 likes -
Replying to @GidMK @Rootclaim
Not sure I follow. Can you give an example of how their result could have been achieved by chance, at a probability >0.001?
1 reply 0 retweets 0 likes
An example: there was underlying selection bias that led to a 'sicker' control group (however we define sicker). Given the small sample size, the logistic model cannot adequately control for this issue, and so the results are hard to interpret
-
-
-
Replying to @saarwilf @Rootclaim
It's pretty complex, but basically once you're into the single digits it's hard to draw much meaning from logistic models i.e.https://bmcmedresmethodol.biomedcentral.com/articles/10.1186/s12874-016-0267-3#Sec12 …
1 reply 0 retweets 0 likes - 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.