General-to-Specific (GETS) and Indicator Saturation (ISAT) Modelling
gets-package.RdThe gets package provides functions and methods for General-to-Specific (GETS) and Indicator Saturation (ISAT) modelling. GETS modelling is a powerful and flexible variable selection algorithm that returns a parsimonious and interpretable model. It is ideally suited for the development of models that can be used for counterfactual and predictive scenario analysis (e.g. conditional forecasting). ISAT modelling provides a comprehensive, flexible and powerful approach to the identification of structural breaks and outliers.
The code of the package originated in relation with the research project G. Sucarrat and A. Escribano (2012). In 2014, Felix Pretis and James Reade joined for the development of the isat code and related functions. Moritz Schwarz and Jonas Kurle joined the development team in 2020.
GETS modelling
GETS methods are available for the following model classes:
Linear regression, both static and dynamic, see
arx,gets.arxandgets.lmVariance models, both static and dynamic, see
arxLogit models, both static and dynamic, see
logitxandgets.logitx
The function arx estimates a static linear regression, or a dynamic AR-X model with (optionally) a log-variance specification. The log-variance specification can either be static or a dynamic log-variance model with covariates (a 'log-ARCH-X' model). For the statistical details of the model, see Section 4 in Pretis, Reade and Sucarrat (2018). The function logitx estimates a static logit model, or a dynamic logit model with covariates (optionally). For complete user-specified GETS modelling, see getsFun.
ISAT modelling
ISAT methods are available for:
Linear regression, both static and dynamic, see
isat
The isat function undertakes GETS model selection of an indicator saturated mean specification. Extraction functions (mainly S3 methods) are also available, together with additional auxiliary functions. For complete user-specified ISAT modelling, see blocksFun.
Vignettes
Two vignettes are available in the package (type browseVignettes("gets") to access them):
An introduction to the gets package
User-Specified General-to-Specific (GETS) and Indicator Saturation (ISAT) Methods
The former is a mildly modified version of Pretis, Reade and Sucarrat (2018), whereas the latter is an updated version of Sucarrat (2020).
Author
| Jonas Kurle: | https://www.jonaskurle.com/ | |
| Felix Pretis: | http://www.felixpretis.org/ | |
| James Reade: | https://sites.google.com/site/jjamesreade/ | |
| Moritz Schwarz: | https://www.inet.ox.ac.uk/people/moritz-schwarz/ | |
| Genaro Sucarrat: | http://www.sucarrat.net/ |
Maintainer: Genaro Sucarrat
References
Jurgen A. Doornik, David F. Hendry, and Felix Pretis (2013): 'Step Indicator Saturation', Oxford Economics Discussion Paper, 658. https://ideas.repec.org/p/oxf/wpaper/658.html
Felix Pretis, James Reade and Genaro Sucarrat (2018): 'Automated General-to-Specific (GETS) Regression Modeling and Indicator Saturation for Outliers and Structural Breaks'. Journal of Statistical Software 86, Number 3, pp. 1-44. doi:10.18637/jss.v086.i03
Carlos Santos, David F. Hendry and Soren Johansen (2007): 'Automatic selection of indicators in a fully saturated regression'. Computational Statistics, vol 23:1, pp.317-335. doi:10.1007/s00180-007-0054-z
Genaro Sucarrat (2020): 'User-Specified General-to-Specific and Indicator Saturation Methods'. The R Journal 12:2, pages 388-401. https://journal.r-project.org/archive/2021/RJ-2021-024/
Genaro Sucarrat and Alvaro Escribano (2012): 'Automated Financial Model Selection: General-to-Specific Modelling of the Mean and Volatility Specifications', Oxford Bulletin of Economics and Statistics 74, Issue 5 (October), pp. 716-735.
Examples
##Simulate from an AR(1):
set.seed(123)
y <- arima.sim(list(ar=0.4), 60)
##Estimate an AR(2) with intercept as mean specification
##and a log-ARCH(4) as log-volatility specification:
myModel <- arx(y, mc=TRUE, ar=1:2, arch=1:4)
##GETS modelling of the mean of myModel:
simpleMean <- getsm(myModel)
#>
#> GUM mean equation:
#>
#> reg.no. keep coef std.error t-stat p-value
#> mconst 1 0 0.026155 0.117029 0.2235 0.82398
#> ar1 2 0 0.333996 0.134355 2.4859 0.01599 *
#> ar2 3 0 0.034490 0.133176 0.2590 0.79661
#> ---
#> Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1
#>
#> GUM log-variance equation:
#>
#> coef std.error t-stat p-value
#> vconst -0.278502 0.391091 0.5071 0.47639
#> arch1 0.215508 0.139456 1.5453 0.12870
#> arch2 0.105365 0.142566 0.7391 0.46340
#> arch3 0.081331 0.144165 0.5642 0.57522
#> arch4 -0.306401 0.140825 -2.1758 0.03443 *
#> ---
#> Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1
#>
#> Diagnostics:
#>
#> Chi-sq df p-value
#> Ljung-Box AR(3) 2.4462 3 0.48509
#> Ljung-Box ARCH(5) 1.3238 5 0.93246
#>
#> 2 path(s) to search
#> Searching:
#> 1
#> 2
#>
#> Path 1: 1 3
#> Path 2: 3 1
#>
#> Terminal models:
#>
#> info(sc) logl n k
#> spec 1 (1-cut): 2.372013 -66.75816 58 1
#>
#> Retained regressors (final model):
#>
#> ar1
##GETS modelling of the log-variance of myModel:
simpleVar <- getsv(myModel)
#> GUM log-variance equation:
#>
#> reg.no. keep coef std.error t-stat p-value
#> vconst 1 1 -0.278502 0.391091 0.5071 0.47639
#> arch1 2 0 0.215508 0.139456 1.5453 0.12870
#> arch2 3 0 0.105365 0.142566 0.7391 0.46340
#> arch3 4 0 0.081331 0.144165 0.5642 0.57522
#> arch4 5 0 -0.306401 0.140825 -2.1758 0.03443 *
#> ---
#> Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1
#>
#> Diagnostics:
#>
#> Chi-sq df p-value
#> Ljung-Box AR(3) 2.4462 3 0.48509
#> Ljung-Box ARCH(5) 1.3238 5 0.93246
#>
#> 3 path(s) to search
#> Searching:
#> 1
#> 2
#> 3
#>
#> Path 1: 2 4 3 5
#> Path 2: 3 4 2 5
#> Path 3: 4 3 2 5
#>
#> Terminal models:
#>
#> info(sc) logl n k
#> spec 1 (1-cut): 2.614267 -66.59622 54 2
#> spec 2: 2.659078 -69.80062 54 1
#>
#> Retained regressors (final model):
#>
#> vconst arch4
##results:
print(simpleMean)
#>
#> Date: Sat Jul 27 15:29:52 2024
#> Dependent var.: y
#> Method: Ordinary Least Squares (OLS)
#> Variance-Covariance: Ordinary
#> No. of observations (mean eq.): 58
#> Sample: 3 to 60
#>
#> SPECIFIC mean equation:
#>
#> coef std.error t-stat p-value
#> ar1 0.34686 0.12327 2.8138 0.006708 **
#> ---
#> Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1
#>
#> SPECIFIC log-variance equation:
#>
#> coef std.error t-stat p-value
#> vconst -0.212822 0.473616 0.2019 0.6532
#> arch1 0.194178 0.142006 1.3674 0.1777
#> arch2 0.070728 0.144627 0.4890 0.6270
#> arch3 0.041949 0.145447 0.2884 0.7742
#> arch4 -0.207758 0.142741 -1.4555 0.1519
#>
#> Diagnostics and fit:
#>
#> Chi-sq df p-value
#> Ljung-Box AR(3) 2.66621 3 0.4460
#> Ljung-Box ARCH(5) 0.50528 5 0.9919
#>
#> SE of regression 0.87620
#> R-squared 0.12105
#> Log-lik.(n=54) -66.75816
print(simpleVar)
#>
#> Date: Sat Jul 27 15:29:52 2024
#> Method: Ordinary Least Squares (OLS)
#> No. of observations (variance eq.): 54
#> Sample: 7 to 60
#>
#> SPECIFIC log-variance equation:
#>
#> coef std.error t-stat p-value
#> vconst -0.71684 0.31600 5.1459 0.02330 *
#> arch4 -0.26115 0.13804 -1.8918 0.06409 .
#> ---
#> Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1
#>
#> Diagnostics and fit:
#>
#> Chi-sq df p-value
#> Ljung-Box AR(3) 2.7674 3 0.4289
#> Ljung-Box ARCH(5) 4.5973 5 0.4670
#>
#> SE of regression 0.88132
#> Log-lik.(n=54) -66.59622
##step indicator saturation of an iid normal series:
set.seed(123)
y <- rnorm(30)
isat(y)
#>
#> SIS block 1 of 2:
#> 15 path(s) to search
#> Searching:
#> 1
#> 2
#> 3
#> 4
#> 5
#> 6
#> 7
#> 8
#> 9
#> 10
#> 11
#> 12
#> 13
#> 14
#> 15
#>
#> SIS block 2 of 2:
#> 14 path(s) to search
#> Searching:
#> 1
#> 2
#> 3
#> 4
#> 5
#> 6
#> 7
#> 8
#> 9
#> 10
#> 11
#> 12
#> 13
#> 14
#>
#> GETS of union of retained SIS variables...
#>
#> GETS of union of ALL retained variables...
#>
#> Date: Sat Jul 27 15:29:52 2024
#> Dependent var.: y
#> Method: Ordinary Least Squares (OLS)
#> Variance-Covariance: Ordinary
#> No. of observations (mean eq.): 30
#> Sample: 1 to 30
#>
#> SPECIFIC mean equation:
#>
#> coef std.error t-stat p-value
#> mconst -0.047104 0.179111 -0.263 0.7944
#>
#> Diagnostics and fit:
#>
#> Chi-sq df p-value
#> Ljung-Box AR(1) 0.046575 1 0.82913
#> Ljung-Box ARCH(1) 3.367118 1 0.06651 .
#> ---
#> Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1
#>
#> SE of regression 0.98103
#> R-squared 0.00000
#> Log-lik.(n=30) -41.49361