frmtmb.survival

Survival families for frmtmb.

• cox() is the proportional hazards model brms calls cox(). Its baseline hazard is an M-spline in time with simplex weights. cox_baseline() gives the fitted weights.
• bhaz() is brms’s addition term for the baseline basis of cox(): time | cens(out) + bhaz(df = 6) ~ x with family = cox() is the model of cox(df = 6).
• royston_parmar() is the flexible parametric model of Royston and Parmar (2002). It writes the log cumulative hazard as a natural cubic spline in log time, parameterized as flexsurv::flexsurvspline() parameterizes it. rp_floored() reports where a fit’s likelihood is a floor and not a density.

Censoring, truncation and random-effect frailties come from the frmtmb formula grammar.

Install

The package is in the frmtmb repository. Install frmtmb first, then this package:

remotes::install_github("aforren1/frmtmb")
remotes::install_github("aforren1/frmtmb",
                        subdir = "extensions/frmtmb.survival")

Example

library(frmtmb.survival)
set.seed(2)
dd <- data.frame(x = rnorm(200))
dd$time <- rexp(200, exp(-0.5 + 0.7 * dd$x))
dd$out <- rbinom(200, 1, 0.3)
fit <- frm(bf(time | cens(out) ~ x), family = cox(), data = dd)
fixef_by_dpar(fit)$mu
cox_baseline(fit)

library(frmtmb.survival) attaches frmtmb too. Without this package, family = "cox" and brmsfamily("cox") stop with a message that names it.