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
|