Don’t calculate post-hoc power, period. This may be a good topic for a thread.https://twitter.com/StatModeling/status/1044214786496462848 …
-
-
Replying to @ADAlthousePhD
All about the CIs. Similarly, would you agree that the prespecified margin of a non-inferiority study is not an important concept post-job? Just look at the upper 95% confidence limit.
2 replies 0 retweets 2 likes -
Replying to @rwyeh
Oh, bobby, it's like throwing chum to a shark. This gives me an excuse to post about my favorite "non-inferiority" example ever: ABSORB-III primary paper results.https://www.ncbi.nlm.nih.gov/pubmed/26457558
1 reply 0 retweets 4 likes -
-
Replying to @ADAlthousePhD @rwyeh
That was a 0.007 for noninferiority (!!!!) of the Absorb scaffold. How? Because the "non-inferiority margin" was an absolute difference of 4.5%
1 reply 0 retweets 3 likes -
Replying to @ADAlthousePhD @rwyeh
Sorry, am I understanding this correctly?: They chose a margin (delta) as an absolute difference of 4.5%. The control (xience stent's) rate was 6.1% - so they were willing to tolerate a value as low as 1.6% (6.1-4.5%) for the intervention, and still say that's within the margins?
1 reply 0 retweets 0 likes -
Other way. As high as ~11% (in the methods, details of power calculation, they list the assumed rates - I think they were anticipating something like 7% event rate w/Xience, so NI margin meant BVS had to rule out event rate above 11.5%)
2 replies 0 retweets 1 like
Although presumably a result of 1.6% would've also been counted as pure noninferiority and not an argument for superiority ...right?
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.