How to interpret Brier score under survival setting?
Opened this issue · 2 comments
Hi, I'm having difficulty understanding the brier score plot stratified by mortality. Why the 0-15 perc has lower values for the short survival times and higher prediciton errors in the long survival times? The trend for 75-100 is a reverse of this observation. In addition, why prediction error is higher at the middle time ranges? See plot attached. Thanks!
?
There is very little data out in tails (larger time) therefore the Brier score (BS) is less likely accurate here. If you consider time <= 20, the BS is lower for the 0-25% group.
Hi @ishwaran, thanks for your reply. However, there is also more data points in time <= 20 for the 75-100% group. why does that group have high (near 0.25, from ~3-18 months) BS scores?