search

chjackson / survextrap

Survival Extrapolation With a Flexible Parametric Model and External Data
https://chjackson.github.io/survextrap/
GNU General Public License v3.0
7 stars 3 forks source link

Allow uncertainty in background hazard #4

Open chjackson opened 1 year ago

chjackson commented 1 year ago

It may be uncertain whether the background hazard is representative of the population being modelled.

The simplest way to address this would be to allow independent prior distributions to be placed on each background hazard value.

Possible syntax survextrap(..., backhaz_sd = "sd") where "sd" (for example) is the name of a column in backhaz containing a prior standard deviation for the background hazard value. Or work on some friendlier scale - suggestions welcome.

A more sophisticated approach would involve a statistical model to explain backhaz as a function of covariates (e.g. as a spline function of age) but that would be a lot of work.

Any approach here should be motivated by some commonly-occurring real situation.

chjackson commented 9 months ago

Comment if you would make use of a feature like this!