ziphsmm (version 2.0.6)

fasthmmfit.cont3: Fast gradient descent algorithm to learn the parameters in a specialized continuous-time zero-inflated hidden Markov model, where zero-inflation only happens in State 1 with covariates in the state-dependent parameters and transition rates.

Description

Fast gradient descent algorithm to learn the parameters in a specialized continuous-time zero-inflated hidden Markov model, where zero-inflation only happens in State 1 with covariates in the state-dependent parameters and transition rates.

Usage

fasthmmfit.cont3(y, x, M, initparm, yceil = NULL, timeindex,
  method = "Nelder-Mead", hessian = FALSE, ...)

Arguments

y

observed time series values

x

matrix of covariates in the state-dependent parameters and transition rates.

M

number of latent states

initparm

vector of initial working parameters for prior, transition, zero proportion, and emission parameters.

yceil

a scalar defining the ceiling of y, above which the values will be truncated. Default to NULL.

timeindex

a vector containing the time points

method

method to be used for direct numeric optimization. See details in the help page for optim() function. Default to Nelder-Mead.

hessian

Logical. Should a numerically differentiated Hessian matrix be returned? Note that the hessian is for the working parameters, which are the generalized logit of prior probabilities (except for state 1), the generalized logit of the transition probability matrix(except 1st column), the logit of non-zero zero proportions, and the log of each state-dependent poisson means

...

Further arguments passed on to the optimization methods

Value

the maximum likelihood estimates of the zero-inflated hidden Markov model

References

Liu, Yu-Ying, et al. "Efficient learning of continuous-time hidden markov models for disease progression." Advances in neural information processing systems. 2015.

Examples

# NOT RUN {
priorparm <- 0
tpmparm <- c(-2,0.1,-0.1,-2,-0.2,0.2)
zeroindex <- c(1,0)
zeroparm <- c(0,-1,1)
emitparm <- c(2,0.5,-0.5,3,0.3,-0.2)
workparm <- c(priorparm,tpmparm,zeroparm,emitparm)
timeindex <- rep(1,1000)
for(i in 2:1000) timeindex[i] <- timeindex[i-1] + sample(1:4,1)

designx <- matrix(rnorm(2000),nrow=1000,ncol=2)
result <- hmmsim3.cont(workparm,2,1000,zeroindex,x=designx,timeindex=timeindex)
y <- result$series
state <- result$state

fit2 <-  fasthmmfit.cont3(y=y,x=designx,M=2,initparm=workparm,
  timeindex=timeindex,
  hessian=FALSE, method="BFGS", control=list(trace=1))
# }