forecast.prob.MSAR: Forecast probabilities for (non) homogeneous MSAR models

In NHMSAR: Non-Homogeneous Markov Switching Autoregressive Models

Description

Computes, for each time t, the conditional probabilities for MSAR models P(Y_t|y_{1:(t-1)}) where Y is the observed process and y the observed time series.

Usage

forecast.prob.MSAR(data, theta, yrange = NULL, covar.emis = NULL, covar.trans = NULL)

Arguments

data	observed time series, array of dimension T*N.samples*d
theta	object of class MSAR including the model's parameter and description. See init.theta.MSAR for more details.
yrange	values at which to compute the forecast probabilities
covar.emis	emission covariate if any.
covar.trans	array of univariate or multivariate series of covariate to take into account in the transition probabilities. The link function is defined in the initialisation step (running init.theta.MSAR.R).

Value

A list containing

..$yrange	abscissa for the forecast probabilities
..$prob	forecast probabilities
Yhat	forecasted value

Author(s)

Valerie Monbet, [e-mail]

See Also

prediction.MSAR

Examples

## Not run
#data(meteo.data)
#data = array(meteo.data$temperature,c(31,41,1)) 
#T = dim(data)[1]
#N.samples = dim(data)[2]
#d = dim(data)[3]
#M = 2
#theta.init = init.theta.MSAR(data,M=M,order=2,label="HH")
#res.hh.2 = fit.MSAR(data,theta.init,verbose=TRUE,MaxIter=200)
#FP = forecast.prob.MSAR(data,res.hh.2$theta)
#plot(data[,1,],typ="l")
#lines(FP$Yhat[,1],col="red")
#alpha = .1
#IC.emp = matrix(0,2,T)
#for (k in 1:length(data[,1,])) {
#	tmp = cumsum(FP$prob[,k,1])/sum(FP$prob[,k,1])
#	IC.emp[1,k] = FP$yrange[max(which(tmp<alpha/2))]
#	IC.emp[2,k] = FP$yrange[max(which(tmp<(1-alpha/2)))]
#}
#lines(IC.emp[1,],lty=2,col="red")
#lines(IC.emp[2,],lty=2,col="red")

Example output

NHMSAR documentation built on Feb. 9, 2022, 9:06 a.m.