Dear all,
We’re preparing a study in which we’d like to use the joint-modelling framework, expecting a somewhat U-shaped effect of the biomarker.
I can’t seem to find any mentioning of this in the documentation or the vignette, but I was wondering if it’s possible to model etavalue
(from the longitudinal model) with a spline in the event model. Building on the first example in ?stan_jm
, I’m thinking something along the lines of this:
library(rstanarm)
library(mgcv)
f1 <- stan_jm(formulaLong = logBili ~ year + (1 | id),
dataLong = pbcLong,
formulaEvent = Surv(futimeYears, death) ~ sex + trt + s(etavalue),
dataEvent = pbcSurv,
time_var = "year",
chains = 1, cores = 1, seed = 12345, iter = 1000)
Cheers,
Ben