2 ms·
The paper we're discussing based its analysis on a bunch of published reports. Eyeballing its supplement Figure 1, about 4/5 of those reports gave a bottom line
by abecedarius 2y ago
The paper we're discussing based its analysis on a bunch of published reports. Eyeballing its supplement Figure 1, about 4/5 of those reports gave a bottom line in the same direction as yours (increased mortality). But the OP paper arrived at a bottom line in the opposite direction.
I haven't looked further at either of these papers; just saying that telling me there's another paper which agrees with 4/5 of the other papers is not actually much information, in itself.
- OutOfHere 2y agoThey're using a very unproven analysis approach, although it's their biased study selection that worries me more. You can prove anything you want if you creatively select mainly the studies that favor your desired conclusions. They didn't even use the 2020 study that I linked. They also didn't use a famous study from 2009 (PMID: 19620159) which showed harm. These are just the two that I could identify quickly.
- abecedarius 2y agoYou may well be right that they got there out of bias; I've only glanced at this. How'd this bias manifest in their choice of which studies to review? The 4/5 opposite that I mentioned points the other way. (Added: I wrote this before you added an example paper.) Of course their main point is supposed to be about finding ways to counter this effect of engineering desired results within academic standards. I agree that's a real thing and part of a huge problem.