EM: EM

In cthmm: Latent Continuous Time Markov Chains

Description

EM algorithm to get MLEs for latent CTMC parameters

Usage

  EM(rates.setup, init.setup, emission.setup, the.data,
    num.subjects, num.states, num.obs.states, tol = 1e-07,
    absorb.state, maxiter = 500, DDO.setup = NULL,
    do.DDO = F, time.dep.emission = F)

Arguments

rates.setup	list with rate setup information
init.setup	list with initial distribution setup information
emission.setup	list with emission distribution setup information
the.data	list with the observed data, with one entry per individual
num.subjects	number of individuals in the study
num.states	number of latent states in the CTMC
num.obs.states	number of observed states
tol	convergence tolerance for the EM
absorb.state	vector of one or more absorbing latent states
maxiter=500	maximum number of iterations of the EM
DDO.setup	setup object for disease driven observation model
do.DDO	Indicator (T/F) if there are disease driven observation times in the model?
time.dep.emission	(Not supported) Indicator if emission distribution has time dependent covariates.

Details

Run the expectation-maximization algorithm to get the maximum likelihood estimates for latent CTMC model parameters for multistate disease models. Data are discretely observed disease processes from multiple independent individuals. Times may be non-informative scheduled times or disease-dependent (informative) times. Uses the SQUAREM package for EM acceleration.

Value

params	MLEs for model parameters: rates/emission/initial/DDO, parameters listed in order of model specification
details	Details of the EM optimization from the SQUAREM package output

.

updates	The iterative oupdates for the parmaeter esitmates at each EM step
LL.updates	Updates of observed data likelihood at each step
time	Run time
num.evals	Number of iterations of the EM
LL	Final complete data log likelihood

Author(s)

Jane Lange

Examples

## Not run: 
library(cthmm)
#load the model setup and example data for a competing risks model
#observed at scheduled and informative observation times.
data(DDO_data)
#run the EM on the example data
DDO_EM=EM(rates.setup=rates.setup,
init.setup=init.setup,
emission.setup=emission.setup,
the.data=DDO_data,
num.subjects=500,
num.states=4,
num.obs.states=3,
tol = 1e-07,
absorb.state=c(3,4),
maxiter = 500,
DDO.setup = DDO.setup,
do.DDO = T)
## End(Not run)

cthmm documentation built on May 2, 2019, 5:46 p.m.