Thank you! It clarifies a lot.
I had time to test more and everything seems nice. The function tve() can be used with multiple variables in the model, giving time varying effects for each covariated specified in this way.
I found out that I had a problem with only one continuous variable - age. Sampling was really slow and sometimes showed an error as follows:
Log probability evaluates to log(0), i.e. negative infinity. Stan can’t start sampling from this initial value.
After several tried it finally compiled the model, giving an output as follows. Does the plot has a meaningful interpretation, i.e. what is the reference used for calculating hazard ratio?
stan_surv
baseline hazard: weibull
formula: Surv(time, status) ~ tve(age)
observations: 210
events: 141 (67.1%)
right censored: 69 (32.9%)
delayed entry: no
------
Median MAD_SD exp(Median)
(Intercept) -8.4 1.1 NA
age 0.1 0.0 1.1
age:tve-bs-coef1 0.0 0.0 1.0
age:tve-bs-coef2 0.0 0.0 1.0
age:tve-bs-coef3 0.0 0.0 1.0
smooth_sd[age] 0.1 0.1 1.1
weibull-shape 0.8 0.1 NA
