StateSpaceModels/localLinearTrend-dlm.R at master · pteetor/StateSpaceModels

This repository has been archived by the owner on Jan 16, 2024. It is now read-only.

# Estimation of the local linear trend model via dlm package

library(dlm)

y <- datasets::Nile

buildModPoly2 <- function(v) {

dV <- exp(v[1])

dW <- exp(v[2:3])

m0 <- v[4:5]

dlmModPoly(order=2, dV=dV, dW=dW, m0=m0)

}

varGuess <- var(diff(y), na.rm=TRUE)

mu0Guess <- as.numeric(y[1])

lambda0Guess <- 0.0

parm <- c(log(varGuess), log(varGuess), log(varGuess),

mu0Guess, lambda0Guess)

mle <- dlmMLE(y, parm=parm, buildModPoly2)

if (mle$convergence != 0) stop(mle$message)

model <- buildModPoly2(mle$par)

filt <- dlmFilter(y, model)

tsdiag(filt)

Provide feedback