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)
|