What do you think about 95% CI presented for many endpoints/ particularly when alpha is split Full video on ZD's Youtube

Dude, ZDogg has lost all validity due to his behavior and rhetoric during COVID and thereafter. He brings your aura down.

lmao was head of CBER and doesn't even know the difference between dual and co-primary endpoints

“The true value exists in this interval 95% of the times”. Not quite. That would be a credible interval not a confidence interval. For a confidence interval, 95% *of the intervals* generated in this way would cover the true value.

@Sensible__Med I was done with ZDoofus in 2021

Get rid of ZD!

How is this relevant in the new world of precision or individualized medicine?

Enjoyed it, totally. Your opinion verges on radical conclusions sometimes. Impact factors is always not everything and only thing in medical publications. These issues are extremely complex at least for a clinician who honestly need to make sense of the study @ihtanboga opinion ?

Vinay Prasad sucks

In observational studies, even a confidence interval that doesn't cross 1 tells you little. With many covariates, the model you pick drives the estimate, and that vibration of effects isn't fixed by alpha correction. Big data makes this worse, not better: n buys you precision, not validity, so you get a precisely wrong answer with a tight interval around it. For multiple endpoints in a trial, alpha correction works. I think Dr. Prasad is talking about RCTs/Clinical trials only.

I won't watch ANYTHING with that clown in it and I'm beginning to doubt your smarts if you're hanging with him.

Hmm

You’re assuming dichotomous endpoints

@Sensible__Med Typically, I state the number of tests, note the Bonferroni threshold p-value, and then present effect sizes, 95% CIs, and raw p-values for findings that are Bonferroni significant. I find this to be the overall most interpretable sequence for readers.

so what was trainees speciality ?