(Total Views: 583)
Posted On: 09/23/2020 12:40:18 PM
Post# of 145890
TechGuru - I have to disagree with your summary of the math.
There are two positive outcomes. The "stop the trial now because it's unfair to the controls" might be set at p<.01, or even p<.005 as you suggest. Then using the "alpha spending" they describe, the final p-value would have to be reduced by that amount (to p<.04 or p<.045 respectively).
The assumption made is that we've received the "you will be significant if current trends continue" message. The question is what penalty we pay to get the details. Say the details show we are at p<.10. That might change whether we ask the FDA for more patients, say - a minor benefit that would only cost a small amount of alpha.
The weird situation is if we ask for details and find out we were between .05 and .01. As I read it, we would not get approval - just continue the trial.
There are two positive outcomes. The "stop the trial now because it's unfair to the controls" might be set at p<.01, or even p<.005 as you suggest. Then using the "alpha spending" they describe, the final p-value would have to be reduced by that amount (to p<.04 or p<.045 respectively).
The assumption made is that we've received the "you will be significant if current trends continue" message. The question is what penalty we pay to get the details. Say the details show we are at p<.10. That might change whether we ask the FDA for more patients, say - a minor benefit that would only cost a small amount of alpha.
The weird situation is if we ask for details and find out we were between .05 and .01. As I read it, we would not get approval - just continue the trial.
![](/m/images/thumb-up.png)
![](/m/images/thumb-down.png)