VAJointSurv (version 0.1.1)

joint_ms_ptr: Creates a joint_ms Object to Estimate a Joint Survival and Marker Model

Description

Creates a joint_ms Object to Estimate a Joint Survival and Marker Model

Usage

joint_ms_ptr(
  markers = list(),
  survival_terms = list(),
  max_threads = 1L,
  quad_rule = NULL,
  cache_expansions = TRUE,
  gh_quad_rule = NULL,
  ders = NULL
)

Value

An object of joint_ms class with the needed C++ and R objects to estimate the model.

Arguments

markers

either an object from marker_term or a list of such objects.

survival_terms

either an object from surv_term or a list of such objects.

max_threads

maximum number of threads to use.

quad_rule

list with nodes and weights for a quadrature rule for the integral from zero to one.

cache_expansions

TRUE if the expansions in the numerical integration in the survival parts of the lower bound should be cached (not recomputed). This requires more memory and may be an advantage particularly with expansions that take longer to compute (like ns_term and bs_term). The computation time may be worse particularly if you use more threads as the CPU cache is not well utilized.

gh_quad_rule

list with two numeric vectors called node and weight with Gauss–Hermite quadrature nodes and weights to handle delayed entry. A low number of quadrature nodes and weights is used when NULL is passed. This seems to work well when delayed entry happens at time with large marginal survival probabilities. The nodes and weights can be obtained e.g. from fastGHQuad::gaussHermiteData.

ders

a list of lists with integer vectors for how the survival outcomes are linked to the markers. 0 implies present values, -1 is integral of, and 1 is the derivative. NULL implies the present value of the random effect for all markers. Note that the number of integer vectors should be equal to the number of markers.

See Also

joint_ms_opt, joint_ms_lb, joint_ms_hess, and joint_ms_start_val.

Examples

# load in the data
library(survival)
data(pbc, package = "survival")

# re-scale by year
pbcseq <- transform(pbcseq, day_use = day / 365.25)
pbc <- transform(pbc, time_use = time / 365.25)

# create the marker terms
m1 <- marker_term(
  log(bili) ~ 1, id = id, data = pbcseq,
  time_fixef = bs_term(day_use, df = 5L),
  time_rng = poly_term(day_use, degree = 1L, raw = TRUE, intercept = TRUE))
m2 <- marker_term(
  albumin ~ 1, id = id, data = pbcseq,
  time_fixef = bs_term(day_use, df = 5L),
  time_rng = poly_term(day_use, degree = 1L, raw = TRUE, intercept = TRUE))

# base knots on observed event times
bs_term_knots <-
  with(pbc, quantile(time_use[status == 2], probs = seq(0, 1, by = .2)))

boundary <- c(bs_term_knots[ c(1, length(bs_term_knots))])
interior <- c(bs_term_knots[-c(1, length(bs_term_knots))])

# create the survival term
s_term <- surv_term(
  Surv(time_use, status == 2) ~ 1, id = id, data = pbc,
  time_fixef = bs_term(time_use, Boundary.knots = boundary, knots = interior))

# create the C++ object to do the fitting
model_ptr <- joint_ms_ptr(
  markers = list(m1, m2), survival_terms = s_term,
  max_threads = 2L)