This page contains (will contain) various GAUSS procedures to adaptively
estimate a variety
of time series and other models. The following is the list of
procedures available along with a
brief description of the contents. The setup and maintenance
of this page is funded by NSF
grant #SBR-970159. These programs are for public noncommercial
use. We make no
performance guarantees.
Primary Authors: Douglas Hodgson, Keith Vorkink, and Irina Solyanik.
Most Recent Update: January, 2002.
Programs
•
ADECM Included are the procedures adecm.g
and jmle.g along with a file
adecm.rdme which discusses installation and the estimation procedure.
The
two .g files adaptively estimate the cointegrating matrix and error correction
matrix in an Error Correction Model. This procedure implements the estimator
discussed in the paper: Hodgson, D. (95), "Adaptive Estimation of
Error
Correction Models," Econometric Theory 14, 44-69. DOWNLOAD
• ADRMA
Included are the procedure adrma.g along with a file adrma.rdme
which discusses installation and the estimation procedure. Adrma.g
procedure
adaptively estimates the cointegrating matrix in the estimation of a cointegrating
regression where the residuals are allowed to follow an ARMA process. This
procedure will implement the estimator discussed in the paper: Hodgson,
D.,
"Adaptive Estimation of Cointegrating Regressions with ARMA Errors,"
Forthcoming in Journal of Econometrics(98). DOWNLOAD
•
STEIG
Included are the procedure steig.g along with a file steig.rdme which
discusses installation and the estimation procedure. Steig.g procedure
adaptively estimates the coefficient vector in the estimation of a linear
regression
model where the residuals are allowed to follow an ARMA process. This
procedure will implement the estimator discussed in the paper: Steigerwald,
D.,(92)
"Adaptive Estimation in Time Series Regression Models," Journal of
Econometrics 54, 251-275. We note this estimation procedure generalizes
two well know models. When no serial dependence exists in the residuals
the
model reduces to Bickel's(82, Annals of Statistics) model.
When no
regressors are present the model reduces to Kreiss' (87, Annals of Statistics)
model. DOWNLOAD
•
ADSUR
Included are the procedure adsur.g with a file adsur.rdme which
discusses installation and the estimation procedure. Adsur.g adaptively
estimates the coefficient vector in the estimation of a Seemingly Unrelated
Regression Model (SUR). The procedure provides estimates of the system
using
GLS, one-step adaptive, and an iterative adaptive estimator. The procedures
implement the estimator discussed in Hodgson, D., O. Linton, and K. Vorkink
(2001), "Testing the Capital Asset Pricing Model Efficiently Under Elliptical
Symmetry: A Semiparametric Approach" forthcoming in Journal of
Applied
Econometrics.
DOWNLOAD
•
ADCAPM
Included are the procedure adcapm.g with a file adcapm.rdme which
discusses installation and the estimation procedure. Adcapm.g adaptively
estimates the
coefficient vector in the estimation of a linear regression model . The
procedure provides
estimates of the parameters using OLS, one-step adaptive, and an iterative
adaptive
estimates. The procedures implement the estiamator discussed in K.
Vorkink(2001),
"Return Distributions and Improved Tests of Asset Pricing Models," working
Paper,
Marriott School of Management. DOWNLOAD
•
WALD Included are the
procedure wald.g along with a file wald.rdme which
discusses installation and the estimation procedure. Wald.g procedure
constructs and performs wald tests. The restrictions must be linear
and both
parameter estimates and covariance matrix of parameters are required inputs
for the procedure. This procedure can be used in conjunction with
the above
estimation procedures. DOWNLOAD
•
DEN
Included are the procedures den.g and syden.g along with a file
density.rdme which discusses installation and the estimation procedure.
Den.g nonparametrically estimates the density of a nxm zero mean series
(m ?=2). Syden.g nonparametrically estimates the density of an nxm
series
assuming symmetry. DOWNLOAD
•
ADARCH
Included are the procedure adarch.g along with a file adarch.rdme which
discusses installation and the estimation procedure . The procedure implements
adaptive estimation of parameters of ARCH model discussed in O. Linton(93),
"Adaptive Estimation in ARCH Models", Econometric Theory, 9, pp.539-569.
Procedure keeps the overall scale parameter of ARCH model fixed and computes
adaptive estimates of regression parameters and identifiable ARCH parameters.
The error density is assumed to be symmetric about zero. DOWNLOAD
•
ADUNIT
Included are the procedure adunit.g, adunit.rdme which discusses installation
and the estimation procedure, and adunit.tex which also discussed the estimation
and
testing procedure. The procedure implements adaptive unit root tests discussed
in
O. Beelders(98), "Adaptive Unit Root Tests", Working Paper, Emory University.
The error density is assumed to be symmetric about zero. DOWNLOAD
•
ADAR
Included are the procedure adar.g, adar.rdme which discusses installation
and the estimation procedure Procedure adar.g is the main procedure
to call
to estimate parameters of AR(p) model. Procedure constructs an estimator
which
is adaptive for all densities of the distribution of the white noise.The
main reference
is Jens-Peter Kreiss(1987), "On Adaptive Estimation In Autoregressive Models
When There Are Nuisance Functions", Statistics & Decisions, 5, pp.
59-76.
DOWNLOAD
•
SEMIPARMA
Included are the procedure semiparma.g, semiparma.rdme which discusses
installation and the estimation procedure. Procedure semiparma.g is the
main
procedure to call to estimate parameters of linear regression model with
ARMA
errors. The main reference is Douglas Hodgson (1998), "Semiparametric Efficient
Estimation in Time Series Regression", University of Rochester Manuscript.
Procedure computes semiparametric estimates of parameters of linear regression
model with ARMA errors in which the innovations to the ARMA process is
stationary and ergodic martingale difference sequence that is also 1st
order Markov
process. Conditional density g(e(t)|e(t-1)) is assumed to be symmetric.
The
semiparametric efficiency bound is also reported. DOWNLOAD
•
ADTEST
Included are the procedure adtest.g along with a
file adtest.rdm, which discusses
installation and the estimation procedure . The procedure constructs
a test statistic for specification tests of conditional heteroscedasticity.
Theoretical
details are discussed in Linton, O., and D. Steigerwald (2000) "Adaptive
Testing
in ARCH Models", Econometric Reviews, 19(2): 146-174.
The semiparametric test statistic is constructed from a nonparametric estimator
of the innovation density and is adaptive (i.e., asymptotically
equivalent to the test
statistic constructed from the true likelihood). The test statistic maximizes
asymptotic local power and weighted average power criteria for the general
family of densities. The asymptotic distribution of test statistic is Gaussian
under
the sequence of local alternatives and is standard Gaussian under the null.
Procedure also constructs the semiparametric estimator of the full parameter
vector and its asymptotic covariance.
DOWNLOAD
*If you need WinZip to extract the above programs, follow the link below
to obtain a copy of the program.
WinZip
Return to Douglas Hodgson's Page
Return
to Keith Vorkink's Page