Keywords - Function groups - @ A B C D E F G H I J K L M N O P Q R S T U V W X Y Z

Library: cafpe
See also: tp/cafpe/hsilv

Quantlet: hoptest
Description: Quantlet for computing the scalar plug-in bandwidth for nonparametric estimation of nonlinear autoregression models of order p. The unknown quantities in the asymptotic optimal bandwidth are nonparametrically estimated using C++ - routines via dlls. Only the gaussian kernel can be used.


Reference(s):

Usage: { Bhat,Bhatr,hB,Chat,sumwc,hC,hA } = hoptest(xsj,yorig,xtj,estimator,kernel,ntotal,sigy2,perB,robden)
Input:
ATTENTION: this quantlet requires to open locling.dll, density.dll (NT) or locling.so, denc.so (UNIX). This can be done with the quantlet cafpeload or directly with garb = dlopen ("\locling.dll") on NT, garb = dlopen ("\density.dll") on NT, garb = dlopen ("/locling.so") on UNIX, garb = dlopen ("/denc.so") on UNIX.
xsj (n x d) matrix of regressors
yorig (ny x 1) vector of dependent data
xtj (ny x d) matrix of lagged variables
estimator character variable for selection nonparametric estimator "loclin": local linear estimator; other estimators not implemented
kernel character variable for selecting kernel for density estimation, must currently be "gaussian"
ntotal scalar, number of observations of available time series
sigy2 scalar, variance of available time series
perB scalar, parameter screening off 0 <= perB < 1 observations with lowest density for estimating B, see also Section 5 in Tschernig & Yang (2000)
robden character variable for switching on/off robustification of density estimation a la Tjostheim & Auestad (1994), see also Section 5 in Tschernig & Yang (2000) "yes": on; "no": off
Output:
Bhat scalar, estimated B, see eq. (3.2) in Tschernig & Yang (2000)
Bhatr scalar, number of observations used to estimate B after screening off perB percent of the observations
hB scalar, predefined bandwidth for estimating B, if set to zero rule-of-thumb bandwidth is computed, see eq. (5.1) in Tschernig & Yang (2000)
Chat scalar, estimated C using partial local quadratic estimation, see Section 5 in Tschernig & Yang (2000)
Chatr scalar, number of observations used to estimate C after screening off perB percent of the observations
hC scalar, predefined bandwidth for estimating C, if set to zero rule-of-thumb bandwidth is computed, see Section 5 in Tschernig & Yang (2000)
hA scalar, plug-in bandwidth for local linear estimation based on estimating the asymptotically optimal bandwidth, see Corollary 2.1 and Section 5 in Tschernig & Yang (2000)

Library: cafpe
See also: tp/cafpe/hsilv

Keywords - Function groups - @ A B C D E F G H I J K L M N O P Q R S T U V W X Y Z

Author: Tschernig 000420
(C) MD*TECH Method and Data Technologies, 27.4.2000