Closed asondhi closed 2 years ago
This should be fixed now in https://github.com/chjackson/flexsurv-dev/commit/ec4f4297dde9312e773ce211d867a8bd87b10ff9 . Thanks for the report.
The issue was that there are two different heuristics that flexsurvspline
uses to determine initial values for optimisation. If the first procedure doesn't work, then it falls back to the second procedure. However when checking if the procedure worked, it only tested for a -log-likelihood of Inf
. In this case, it gave a log-likelihood of NaN
. It now checks for NaN
as well.
wow thank you so much for the quick fix!! :) much appreciated!
When fitting regression models given left truncated survival data, and using the truncation time as a regressor (for example, to test for dependent LT flexibly),
flexsurvspline
tends to have issues. Example below gives the error: