Hi, Tech, We agree that the Humanigen DSMB chan
Post# of 155138
(Total Views: 460)
Posted On: 02/25/2021 4:29:54 PM
Hi, Tech,
We agree that the Humanigen DSMB changes are the test case.
It is true that the DSMB won't recommend the more patients if the results-to-date aren't in the promising zone. However, that doesn't mean they use the current interim improvements (as per the formula you provide for post-hoc power) in that calculation. They use the original target improvement - presumably 30%. That's conceptually and numerically very different.
Sure, if you have great data to date, the remaining observations don't have to be so great to reach the target. That's subtraction, and how they figured they were at 1.37 at interim.
My last point was I'm pretty sure that "P=.05" and "post-hoc power = 50%" are the same event. You can check it out with our boundary significance cases. I'll admit I haven't run these - just arguing symmetry.
Rest of Board - hope this is not too distracting. Tech - happy to continue this geekiness in private chat.
Thanks!
We agree that the Humanigen DSMB changes are the test case.
It is true that the DSMB won't recommend the more patients if the results-to-date aren't in the promising zone. However, that doesn't mean they use the current interim improvements (as per the formula you provide for post-hoc power) in that calculation. They use the original target improvement - presumably 30%. That's conceptually and numerically very different.
Sure, if you have great data to date, the remaining observations don't have to be so great to reach the target. That's subtraction, and how they figured they were at 1.37 at interim.
My last point was I'm pretty sure that "P=.05" and "post-hoc power = 50%" are the same event. You can check it out with our boundary significance cases. I'll admit I haven't run these - just arguing symmetry.
Rest of Board - hope this is not too distracting. Tech - happy to continue this geekiness in private chat.
Thanks!
(2)
(0)