Replies to post #5460 on Dendreon Corp fka DNDNQ

[p3analyze]

Lest me not forget the beauty of the haybittle-peto (which is essentially that of Armitage), the closer the interim and final analyses are spaced, the higher the alpha available for the final. For example, a 180 of 360 split, the interim of alpha 0.01 would mean there is about 0.046 available at final. Using the same interim alpha of 0.01, at a 210 of 360 split, the final alpha would be 0.0474. Using an interim alpha of 0.005, at the 210 fo 360 split, the final alpha would be 0.0493.

[iwfal]

Ocyan - My point was that no matter which spend function they moved from and which spend function they moved to it is probably impossible to justify the decrease in trial size for the small increase in final alpha once you take into account that much of the power of the trial is likely to be in the tail.