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/fpenps

Quantlet: fpenpsl
Description: Quantlet to compute lag selection criteria for nonlinear autoregressive models for a given vector of lags. It allows to compute two criteria based on local linear estimation of the Asymptotic Final Prediction Error: AFPE and CAFPE. If a scalar bandwidth is given, it is used as hA in the computation of AFPE and CAFPE. If a vector bandwidth is given, only the residuals are computed and zeros returned for the criteria.

Reference(s):

Usage: { critoone,Ahat,sig2wnhat,hA,hB,hC,critgrid,xtj,yorig,resid } = fpenpsl(xraw,xresid,lags,lagmax,hAgiven,volat,startval,robden,estimator,kernel,selcrit,perA,perB,Ksqint)
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.
xraw (n x 1) vector of univariate time series
xresid (nr x 1) vector of residuals for selecting lags of conditional volatility function; if not needed set xresid = 0
lags (d x 1) vector of lags
lagmax scalar, largest lag to be considered
hAgiven scalar bandwidth for which if set to zero a scalar bandwidth is computed using hoptest or (d x 1) vector of bandwidth for which only the residuals are computed but no critiera.
volat "no": lag selection for conditional mean function; "resid": lag selection for conditional volatility function, the residuals of fitting a conditional mean function have to be contained in xresid
startval character variable, to control treatment of starting values "different": for each lag vector as few starting values are used as necessary; "same": for each lag vector the same starting value is used which is determined by the largest lag used in the lag selection quantlet xorigxe
robden character variable for switching on/off robustification of density estimation a la Tjostheim & Auestad (1994), see also Section 5 in TY "yes": on; "no": off
estimator character variable for selection nonparametric estimator "loclin": local linear estimator; other estimators not implemented
kernel character variable, kernel used; "gaussian": Gaussian kernel
selcrit character variable to select lag selection critierion: "lqafpe": estimating the asymptotic Final Prediction Error (AFPE) using local linear estimation and a plug-in bandwidth based on partial local quadratic estimator "lqcafpe": estimating the corrected asymptotic Final Prediction Error (CAFPE) using local linear estimation and a plug-in bandwidth based on partial local quadratic estimator
perA scalar, parameter used for screening off 0 <= perA <= 1 percent of the observations with the lowest density for estimating A, see eq. (3.1) and Section 5 in TY
perB scalar, parameter like perA but for screening off perB observations with lowest density for estimating B
Ksqint scalar, constant of kernel ||K||_2^2 = integral of K(u)^2 du
Output:
critone scalar, estimated criterion value (requires scalar bandwidth)
Ahat scalar, computed Ahat of TY, eq. (3.1)
sig2wnhat scalar, estimated variance of white noise based on local linear estimation, see TY, eq. (3.4) (requires scalar bandwidth)
hA scalar, estimated asymptotically optimal bandwidth for estimating A and (C)AFE
hB scalar, bandwidth for estimating B which is used for plug-in bandwidth hA and (C)AFPE
hC scalar, bandwidth for estimating C which is used for plug-in bandwidth hA
critgrid vector, estimated criteria AFPE and CAFPE for a given lag vector (requires scalar bandwidth)
xtj (nr x d) matrix of lagged variables
yorig (nr x 1) vector of dependent observations
resid (nr x 1) vector of residuals which are estimated for the given lag vector

Library: cafpe
See also: tp/cafpe/fpenps

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