Replies to post #138558 on Biotech Values

[iwfal]

ONTY

There's no data on interims in the calculation whatsoever. You put in the powering assumptions (90%, 20 vs 26, 2:1, .025 one-sided) and it tells you you need 705 events at the final look.

No. You have to put in .045 as the two sided alpha to get 705. If the interims were as small as claimed by Wedbush the alpha you'd have to put in to get 705 events would be over 0.049. And if they were as small as you thought it would be just shy of 0.0500.

[p3analyze]

>> Where is the impact of the interim structure on this outcome?<<
A multivariate normal function (one variate for each interim analysis) integrated over the interval set by the OBF type critical values would accomplish just that. I think that's what Clark meant by his function.