ONTY - Interesting reverse engineer of fact that they did, in fact, use something close enough to OBF as makes no difference:
a) We know they 90% powered based upon a 6 month survival advantage over 20 month SOC. Equating to HR=0.7692.
b) That 90% powering end up with a trigger of 705 events.
c) We know that they gave out 0.025 as the total alpha (single sided - or 0.05 double sided).
d) Given that almost all the power is in the last look - and this would be even more true in a interim methodology more conservative than OBF - then we can calculate the final alpha.
...
Final alpha is 0.045 if there were no interims. There is a missing 0.004, undoubtedly alotted to the interims. Obviously you can then iterate on this - alotting some alpha to the interims, recalculating 90% power, ... . In the end it is crystal clear that the alpha assigned to the final look was in the range of 0.043-0.044. Which means a fairly healthy alpha was assigned to the interims - almost precisely the size you would expect for OBF.
NOTE: Obviously I realize that this will convince none of the kool-aid drinking longs (who will just argue that it is coincidence it works out so close - probably arguing that they used a different HR, but note that even the best powering HR they have ever given produces about the same results. And the worst powering HR they have given... the longs wouldn't want to go there.). And of course most of this board will consider my calc to be thus a waste of time - but I find it fun to reverse engineer around the arbitrary and, IMO, immoral hiding of data. (Immoral because it allows longs to dupe themselves - although there is a case that, ultimately, you cannot protect some people from themselves.)
AI
