[R-pkgs] Version 0.7 of package tsDyn, nonlinear time series

Matthieu Stigler matthieu.stigler at gmail.com
Sun Jul 26 16:07:06 CEST 2009


Version 0.7 of package tsDyn presented at useR! 2009  is now on CRAN, 
extended with several new features.

The package tsDyn is aimed at estimating nonlinear time series models 
which exhibit regime specific properties. The regime switching dynamics 
can either be described by smooth transition (STAR and LSTAR) or 
threshold effects (SETAR). The package furthermore offers nonlinear 
models such as neural networks (NNET), and additive autoregressive (AAR) 

The version 0.7 enhances the functionalities for the threshold 
autoregression models (SETAR) by extending the estimation techniques, 
providing a complete testing framework and finally allowing its use in a 
multivariate framework (generalizing VAR and VECM models).

Those new features are particularly interesting in economic applications 
as they generalize the concept of cointegration: with threshold 
cointegration, variables are still meant to share a long-run 
relationship, but their adjustment need not occur instantaneously but 
only after the deviation exceeds some critical threshold. This allows 
the model to take into account possible effects of transaction costs or 
stickiness of prices. Further, it permits us to capture asymmetries in 
the adjustment process, where positive or negative deviations are not 
corrected to the same extent.

A second vignette available at 
http://code.google.com/p/tsdyn/wiki/ThresholdCointegration offers an 
overview of the threshold cointegration framework and describes the 
implemented functions. It can serve as an ideal introduction for people 
interested in discovering this field.

Main new features include:
-added possibilty to have two thresolds and hence three regimes in setar 
and selectSETAR (arg nthresh)
-new functions for unit roots tests: KapShinTest() and BBCTest()
-new functions for estimating VAR and VECM: lineVar
-new function for estimating TVECM: function TVECM()
-new function for estimating TVAR: function TVAR()
-new function to test for setar: function setarTest()
-new function to test for TVAR: function TVAR.LRtest()
-new function to test for TVECM: functions TVECM.SeoTest() and 
-new function to simulate/bootstrap a TVAR: function TVAR.sim()
-new function to simulate/bootstrap a TVECM: function TVECM.sim()
-new function to simulate/bootstrap a setar: function setar.sim()
-new function to estimate regime-specific variance in setar: function 
-new function to extend a bootstrap replication in setarTest: function 
-added in selectSETAR() and setar() following args: include, common, 
model, trim, MM, ML, MH, model, restriction
-added in selectSETAR(): criterion "SSR" (sum of squares residual) and 
argument max.iter
-extended arg th in selectSETAR to search inside an intervall or around 
a point or on the whole grid

Note that minor fixes/improvements are expected soon, those will be only 
reported in the specific mailing list:  tsdyn at googlegroups.com

Any comments/remarks, suggestions and bug reports are sure welcome!

Matthieu Stigler

More information about the R-packages mailing list