Hacker News new | ask | show | jobs
by greenhatman 1754 days ago
>all of which apparently lose all statistically-significant support for ivermectin once the fraudulent Egyptian study is removed from the data set

That's not my understanding from reading here: https://ivmmeta.com/

Looks at the graphs there, and then check out a few of the studies with low p-values.
1 comments
gus_massa 1754 days ago
If you go to https://ivmmeta.com/ I suggest to ignore all the non RCT studies because they are usually very bad.

There are only 31 RTC studies. Most of them are very small and not even statically significant. It is a known problem that studies that show no effect or a bad effect are never published due to report and publication bias. So it's better to ignore the not statically significant studies.

So there are only 6 of the 31 RTC studies that are statically significant. If Ivermectin has no effect, we expect to see like 1.5 studies that are "statistically significant" due to a flukes. We don't know how many other studies have been tried but never published, so let's be very conservative and assume that if Ivermectin has no effect, we would get 2 studies that are "statistically significant" due to a flukes.

There are 6. I read a few of them and have very strange things. Not smoking guns, but big red flags.

So ... can you choose your favorite 3 statistically significant? I'll dismiss 2 as flukes and hopefully I can explain why the other study has a very big red flag and is not reliable.

link