Usage: |
{ mA,gsqA,denA,err } = fvllc(Xsj,Yorig,h,Xtj,kernreg,lorq,fandg,loo);
|
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 ("/locling.so") on UNIX.
|
| Xsj | (n x d) matrix of regressors
|
| Yorig | (ny x 1) vector of dependent data
|
| h | scalar or (d x 1) vector, bandwidth
|
| Xtj | (nxest x d) matrix of data at which
estimation is conducted
|
| kernreg | must currently be "gaussian"
|
| lorq | 1=local linear, 2=local quadratic with direct second derivatives
|
| fandg | 0=only mean function, 1=also volatility function
(be careful: volatility function may not be positive)
|
| loo | 0=take all, 1 = leave-one-out (only works if n=ny)
|
Output: |
| mA | (nxest x (lorq*d+1)) matrix of
cond. mean, cond. first der., cond. direct
sec. der. (lorq=2)
|
| gsqA | (nxest x (lorq*d+1)) matrix of
cond. volatility, cond. first der.,
cond. direct sec. der. (lorq=2)
|
| denA | (nxest x 1) vector of cond. density
|
| err | 0 if matrix inversion in dll ok, 1 if matrix inversion failed
|