(Total Views: 709)
Posted On: 01/07/2021 4:17:43 PM
Post# of 153988
Thanks for the updates - appreciate your analysis and format. A few thoughts:
1) Interesting that the magic number for statistical significance (a bit more than a 30% reduction) is the same number cited by Hahn (right person?) as his target for great clinical significance. (Stat power underpins all?)
2) We've been thinking in that direction - what percentage of deaths can we prevent - and hoping there were enough events that that percentage was stat sig.
3) It may be time to "flip the script" on the direction of our analysis - start by re-estimating deaths in the control group and subtracting to get treatment deaths. I've been using a 28% to 35% range - about 37- to 46+ deaths.
There is new data today: 35.8% deaths in control group of the 4k person Roche/Sanofi trial. That would be 47 control deaths in our trial, leaving only 40 left over treatment deaths. How many zeros in the p-value?
1) Interesting that the magic number for statistical significance (a bit more than a 30% reduction) is the same number cited by Hahn (right person?) as his target for great clinical significance. (Stat power underpins all?)
2) We've been thinking in that direction - what percentage of deaths can we prevent - and hoping there were enough events that that percentage was stat sig.
3) It may be time to "flip the script" on the direction of our analysis - start by re-estimating deaths in the control group and subtracting to get treatment deaths. I've been using a 28% to 35% range - about 37- to 46+ deaths.
There is new data today: 35.8% deaths in control group of the 4k person Roche/Sanofi trial. That would be 47 control deaths in our trial, leaving only 40 left over treatment deaths. How many zeros in the p-value?
(11)
(0)
