It's not the inclusion criteria, which are basically "chuck all the awful studies into one website". It's just that the authors extract only "positive" results regardless of whether studies actually showed a benefit
-
-
Replying to @GidMK @K_Sheldrick
You keep conflating "didn't show a benefit" with "wasn't statistically significant", but the latter doesn't make a difference in a binary p-value computation.
2 replies 1 retweet 34 likes -
And your idea of "number wasn't corrected properly" means "awful study" is probably only slightly correlating in reality. I've read a 5-digit number of studies, and I find such mistakes in the majority of them, at every level of clinical quality.
1 reply 0 retweets 32 likes -
Replying to @EduEngineer @K_Sheldrick
Yes as I said there are a lot of awful studies. This one is particularly worthless tho, for a number of reasons not limited to the one that you want to ignore for some reason
1 reply 0 retweets 1 like -
Replying to @GidMK @K_Sheldrick
I'm not ignoring it. I stated that it was worth contacting the author and figuring it out, but before that step, it's not a reason to suggest "worthless".
2 replies 1 retweet 31 likes -
Replying to @EduEngineer @K_Sheldrick
That's one element of quality. It's not even the only thing that I pointed out in this thread. Other issues are the lack of reporting of potential confounders, the inadequate control group, the lack of reporting generally, the contradictions btwn pre-reg and study...
2 replies 0 retweets 6 likes -
Replying to @GidMK @K_Sheldrick
Unless there is a reason to believe confounders skew one direction, magically, missed confounders are understood not to have a significant effect on such computed p-values. In fact, that's the point of a p-value: it only asks the question of whether the results happen at random.
2 replies 0 retweets 26 likes -
Replying to @EduEngineer @K_Sheldrick
By definition a confounder skews in one direction, otherwise it does not confound the relationship and is not a confounder
3 replies 0 retweets 13 likes -
Replying to @GidMK @K_Sheldrick
A confounder skews in one direction *internally* to a single study, but those wash out at high numbers of studies/coin flips. What you need are global confounders to make the argument.
3 replies 0 retweets 32 likes -
Replying to @EduEngineer @GidMK
Gid, there's a point at which you have to accept that some anon pulling out random statistical words and claiming to be a statistician... isn't. I'm not spending more of my Saturday on this.
3 replies 0 retweets 7 likes
Yeh fair. It's wild how they keep saying "but Fisher!" in response to critique, as if that meant anything. So weird
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.