markov_sde: Predictions for continuous time, nonhomogeneous Markov multi-state models using Aalen's additive hazards models.

Description

A numerically efficient algorithm to calculate predictions from a continuous time, nonhomogeneous Markov multi-state model. The main inputs are are a list of Aalen's additive hazards models, the initial values, the transition matrix and the covariate patterns. The predictions include state occupancy probabilities and length of stay. Standard errors are calculated using the delta method. Includes differences and standardisation.

Usage

markov_sde(models, trans, newdata, init = NULL, nLebesgue = 10000 + 1, los = FALSE,
           nOut = 300, weights = 1)
# S3 method for markov_sde
standardise(x, ...)
# S3 method for markov_sde
plot(x, y, stacked=TRUE, which=c("P","L"), index=NULL,
    xlab="Time", ylab=NULL, col=2:6, border=col,
    ggplot2=FALSE, lattice=FALSE, alpha=0.2,
    strata=NULL,
    ...)
# S3 method for markov_sde
as.data.frame(x, row.names=NULL, optional=NULL, ci=TRUE,
    P.conf.type="logit", L.conf.type="log",
    P.range=c(0,1), L.range=c(0,Inf),
    ...)

Value

markov_sde returns an object of class

"markov_sde".

Arguments

models: list of models. Currently allows only for aalen regression models.
trans: Transition matrix describing the states and transitions in the multi-state model. If S is the number of states in the multi-state model, trans should be an S x S matrix, with (i,j)-element a positive integer if a transition from i to j is possible in the multi-state model, NA otherwise. In particular, all diagonal elements should be NA. The integers indicating the possible transitions in the multi-state model should be sequentially numbered, 1,...,K, with K the number of transitions. See msprep
newdata: data.frame of the covariates to use in the predictions
init: vector of the initial values with the same length as the number of states. Defaults to the first state having an initial value of 1 (i.e. "[<-"(rep(0,nrow(trans)),1,1)).
nLebesgue: Number of steps for the continuous integration
los: logical variable for whether to estimate the length of stay
nOut: number of rows to represent the continuous changes
weights: numeric vector to represent differences or standardisation

For plot.markov_sde:

y: (currently ignored)
stacked: logical for whether to stack the plots. Default: TRUE
index: indicator of which row of newdata to plot
which: character to indicate either transition probabilities ("P") or length of stay ("L")). Default: "P".
xlab: x-axis label
ylab: x-axis label
col: colours (ignored if ggplot2=TRUE)
border: border colours for the polygon (ignored if ggplot=TRUE)
ggplot2: use ggplot2
alpha: alpha value for confidence bands (ggplot)
lattice: use lattice
strata: formula for the stratification factors for the plot

For as.data.frame.markov_sde:

row.names: add in row names to the output data-frame
optional: (not currently used)
ci: logical for whether to include confidence intervals. Default: TRUE
P.conf.type: type of transformation for the confidence interval calculation for the state occupancy probabilities. Default: logit transformation. This is changed to "identity" if any of the weights are negative
L.conf.type: type of transformation for the confidence interval calculation for the length of stay calculation. Default: log transformation. "identity" if any of the weights are negative
P.range: valid values for the state occupancy probabilities. Default: (0,1).
L.range: valid values for the state occupancy probabilities. Default: (0,Inf).

For standardise.markov_sde:

x: object to extract standardised values
...: other arguments. For plot.markov_sde, these arguments are passed to plot.default. For standardise.markov_sde, these arguments are not used, as the standardisation must be done earlier in markov_sde.

Author

Mark Clements

Details

Uses an approach developed by Ryalen and colleagues. This is a re-implementation in C++.