With the R-package
hystar
, you can simulate
data from the hysteretic threshold autoregressive (HysTAR) model, and
estimate its parameters. It comes with three functions:
-
hystar_fit
, to estimate the HysTAR parameters with the conditional least squares method, using your own data or simulated data, -
z_sim
, to simulate a threshold variable, -
hystar_sim
, to simulate an outcome variable.
Results from the time series analysis can be assessed with the standard
methods in R, like plot
, summary
and print
. Additionally, you can
extract the predictive residuals with the residuals
-method for further
analysis.
A minimal example:
library(hystar)
#>
#> __ __
#> / /_ __ ______/ /_________
#> / _ / // (_ -/ _/ _ / __\
#> /_//_/\_, /___)\__/\_,_/_/
#> /___/ 1.2.0
#>
#> Estimation and simulation of the HysTAR Model.
#> For function help, run `?hystar_fit`, `?hystar_sim` or `?z_sim`.
#> For more information, run `hystar_info()` (opens a URL in your browser).
control_variable <- z_sim(n_t = 100)
simulated_hystar_model <- hystar_sim(z = control_variable)
fitted_hystar_model <- hystar_fit(data = simulated_hystar_model$data)
summary(fitted_hystar_model)
#> HysTAR model fitted on 99 observations, of which
#> 51 observations in regime 0 and
#> 48 observations in regime 1.
#>
#> Estimated thresholds:
#> r0 r1
#> -0.454 0.562
#>
#> Estimated delay:
#> 0
#>
#> Estimated model coefficients:
#> est SE p
#> phi_00 0.314 0.156 0.045
#> phi_01 0.346 0.106 0.001
#> phi_10 1.882 0.435 0.000
#> phi_11 0.538 0.108 0.000
#>
#> Estimated residual variances:
#> sigma2_0 sigma2_1
#> 1.009 1.097
#>
#> Residuals:
#> min 1q median 3q max
#> -2.639 -0.676 0.014 0.823 2.532
#>
#> Information criteria:
#> bic aic aicc aiccp
#> 28.28185 16.87277 17.92886 28.87277
For the current CRAN release (1.0.0):
install.packages("hystar")
For the development version (1.2.0.9000):
devtools::install_github("daandejongen/hystar")
If you have used this package for an scientific publication, please cite it with:
De Jong, D. (2022). hystar: Simulation and Estimation of the Hysteretic TAR Model. R package version 1.2.0, https://github.com/daandejongen/hystar/.
BibTeX:
@Manual{,
title = {hystar: Simulation and Estimation of the Hysteretic TAR Model},
author = {Daan {de Jong}},
year = {2022},
note = {R package version 1.2.0},
url = {https://github.com/daandejongen/hystar/},
}
For more information about the package, see the hystar
website.
If you want to read more about the HysTAR model itself, see the paper with the original proposal of the HysTAR model in Biometrika (Li, Guan, Li and Yu (2015)). Or, for a mathematically more accessible introduction, see the paper (pre-print) I wrote about detecting hysteresis with the HysTAR model in psychological time series.