pmatrix.simfs: Transition probability matrix from a fully-parametric,...

Description Usage Arguments Details Value Author(s) See Also Examples

View source: R/mstate.R

Description

The transition probability matrix for semi-Markov multi-state models fitted to time-to-event data with flexsurvreg. This has r,s entry giving the probability that an individual is in state s at time t, given they are in state r at time 0.

Usage

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
pmatrix.simfs(
  x,
  trans,
  t = 1,
  newdata = NULL,
  ci = FALSE,
  tvar = "trans",
  tcovs = NULL,
  M = 1e+05,
  B = 1000,
  cl = 0.95,
  cores = NULL
)

Arguments

x

A model fitted with flexsurvreg. See msfit.flexsurvreg for the required form of the model and the data. Additionally this should be semi-Markov, so that the time variable represents the time since the last transition. In other words the response should be of the form Surv(time,status). See the package vignette for further explanation.

x can also be a list of models, with one component for each permitted transition, as illustrated in msfit.flexsurvreg.

trans

Matrix indicating allowed transitions. See msfit.flexsurvreg.

t

Time to predict state occupancy probabilities for. This must be a single number, unlike pmatrix.fs.

newdata

A data frame specifying the values of covariates in the fitted model, other than the transition number. See msfit.flexsurvreg.

ci

Return a confidence interval calculated by simulating from the asymptotic normal distribution of the maximum likelihood estimates. This is turned off by default, since two levels of simulation are required. If turned on, users should adjust B and/or M until the results reach the desired precision. The simulation over M is generally vectorised, therefore increasing B is usually more expensive than increasing M.

tvar

Variable in the data representing the transition type. Not required if x is a list of models.

tcovs

Predictable time-dependent covariates such as age, see sim.fmsm.

M

Number of individuals to simulate in order to approximate the transition probabilities. Users should adjust this to obtain the required precision.

B

Number of simulations from the normal asymptotic distribution used to calculate confidence limits. Decrease for greater speed at the expense of accuracy.

cl

Width of symmetric confidence intervals, relative to 1.

cores

Number of processor cores used when calculating confidence limits bu repeated simulation. The default uses single-core processing.

Details

This is computed by simulating a large number of individuals M using the maximum likelihood estimates of the fitted model and the function sim.fmsm. Therefore this requires a random sampling function for the parametric survival model to be available: see the "Details" section of sim.fmsm. This will be available for all built-in distributions, though users may need to write this for custom models.

Note the random sampling method for flexsurvspline models is currently very inefficient, so that looping over the M individuals will be very slow.

pmatrix.fs is a more efficient method based on solving the Kolmogorov forward equation numerically, which requires the multi-state model to be Markov. No error or warning is given if running pmatrix.simfs with a Markov model, but this is still invalid.

Value

The transition probability matrix. If ci=TRUE, there are attributes "lower" and "upper" giving matrices of the corresponding confidence limits. These are formatted for printing but may be extracted using attr().

Author(s)

Christopher Jackson chris.jackson@mrc-bsu.cam.ac.uk.

See Also

pmatrix.fs,sim.fmsm,totlos.simfs, msfit.flexsurvreg.

Examples

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
# BOS example in vignette, and in msfit.flexsurvreg

bexp <- flexsurvreg(Surv(years, status) ~ trans, data=bosms3, dist="exp")
tmat <- rbind(c(NA,1,2),c(NA,NA,3),c(NA,NA,NA))

# more likely to be dead (state 3) as time moves on, or if start with
# BOS (state 2)

pmatrix.simfs(bexp, t=5, trans=tmat)
pmatrix.simfs(bexp, t=10, trans=tmat)

# these results should converge to those in help(pmatrix.fs), as M
# increases here and ODE solving precision increases there, since with
# an exponential distribution, the semi-Markov model is the same as the
# Markov model.

flexsurv documentation built on Sept. 13, 2021, 5:07 p.m.