Kabunshi, censored is any patient that does not have a recorded death. So the total will at all times be N - events.
The problem with his curves are WHEN the censors happened. With very few exceptions, censors will be those who are still alive. The 100.
It should be fairly obvious that these will be kicking in hard at about 20 months, then slowly falling off till about 5 years. Perhaps one or two at the 8-9 year mark for the early enrollees.
Yet his data shows nothing at all like that. His treatment arm has all the censors clustered at 27-35 months (i.e, nobody enrolled prior to Summer '14 is alive). While the control arm is all clustered at 105 months (i.e., 21 patient on control have lived 9 years).
This just makes no sense at all.
It looks to me that he is simply filling in a K/M table to match what he thinks the results should be. But the K/M curves ARE the results, so there is not really a model here at all. Just a prediction, then a numerical analysis of that.
News 
Market Data 
Discover 
AI
