expandPar: Expand vector of free working parameters to vector of all...

In bmcclintock/momentuHMM: Maximum Likelihood Analysis of Animal Movement Behavior Using Multivariate Hidden Markov Models

Expand vector of free working parameters to vector of all working parameters including any fixed parameters (used in fitHMM.R and nLogLike.R)

Description

Expand vector of free working parameters to vector of all working parameters including any fixed parameters (used in fitHMM.R and nLogLike.R)

Usage

expandPar(
  optPar,
  optInd,
  fixPar,
  wparIndex,
  betaCons,
  deltaCons,
  nbStates,
  nbCovsDelta,
  stationary,
  nbCovs,
  nbRecovs = 0,
  mixtures = 1,
  nbCovsPi = 0
)

Arguments

optPar	vector of free working parameters
optInd	indices of constrained parameters
fixPar	Vector of working parameters which are assumed known prior to fitting the model (NA indicates parameters is to be estimated)
wparIndex	Vector of indices for the elements of fixPar that are not NA
betaCons	Matrix of the same dimension as beta0 composed of integers identifying any equality constraints among the t.p.m. parameters.
deltaCons	Matrix of the same dimension as delta0 composed of integers identifying any equality constraints among the initial distribution working scale parameters.
nbStates	Number of states of the HMM
nbCovsDelta	Number of initial distribution covariates
stationary	FALSE if there are time-varying covariates in formula or any covariates in formulaDelta. If TRUE, the initial distribution is considered equal to the stationary distribution. Default: FALSE.
nbCovs	Number of t.p.m. covariates
nbRecovs	Number of recharge covariates
mixtures	Number of mixtures for the state transition probabilities
nbCovsPi	Number of mixture probability covariates

Value

A vector of all working parameters including any fixed parameters

Examples

## Not run: 
nbStates <- 2
stepDist <- "gamma" # step distribution
angleDist <- "vm" # turning angle distribution

# extract data from momentuHMM example
data <- example$m$data

### 1. fit the model to the simulated data
# define initial values for the parameters
mu0 <- c(20,70)
sigma0 <- c(10,30)
kappa0 <- c(1,1)
stepPar <- c(mu0,sigma0) # no zero-inflation, so no zero-mass included
anglePar <- kappa0 # not estimating angle mean, so not included
formula <- ~cov1+cos(cov2)

# constrain cov1 effect to state 1 -> 2 and cov2 effect to state 2 -> 1
fixPar <- list(beta=c(NA,NA,0,NA,0,NA))

m <- fitHMM(data=data,nbStates=nbStates,dist=list(step=stepDist,angle=angleDist),
            Par0=list(step=stepPar,angle=anglePar),formula=formula,fixPar=fixPar)

# convert free parameter vector (m$mod$wpar) to full set of working parameters (m$mod$estimate)
est <- momentuHMM:::expandPar(m$mod$wpar,m$conditions$optInd,unlist(m$conditions$fixPar),
                              m$conditions$wparIndex,m$conditions$betaCons,m$conditions$deltaCons,
                              nbStates,
                              ncol(m$covsDelta)-1,m$conditions$stationary,nrow(m$mle$beta)-1)

all(est==m$mod$estimate)

## End(Not run)

bmcclintock/momentuHMM documentation built on Oct. 26, 2022, 1 a.m.