1% Generated by roxygen2: do not edit by hand 2% Please edit documentation in R/artransform.R 3\name{artransform} 4\alias{artransform} 5\title{Mapping real valued parameters to stationary region} 6\usage{ 7artransform(param) 8} 9\arguments{ 10\item{param}{Real valued parameters for the transformation.} 11} 12\value{ 13transformed The parameters satisfying the stationary constrains. 14} 15\description{ 16Function \code{artransform} transforms \eqn{p} real valued parameters to 17stationary region of \eqn{p}th order autoregressive process using 18parametrization suggested by Jones (1980). Fortran code is a converted from 19\code{stats} package's C-function \code{partrans}. 20} 21\note{ 22This should in theory always work, but in practice the initial transformation 23by \code{tanh} can produce values numerically identical to 1, leading to AR coefficients 24which do not satisfy the stationarity constraints. See example in \code{logLik.SSModel} on how 25to scope with those issues. 26} 27\examples{ 28artransform(1:3) 29} 30\references{ 31Jones, R. H (1980). Maximum likelihood fitting 32of ARMA models to time series with missing observations, Technometrics 33Vol 22. p. 389--395. 34} 35