Usage: |
npfit = npgarchest(y, {h1 {,h2 {,N {,R {,debug}}}}})
|
Input: |
| y | n x 1 vector, time series.
|
| h1 | optional scalar, bandwidth for univariate density.
(If h1 <= 0 or not given, it is set to
standard deviation of 4/log(n))
|
| h2 | optional 2 x 1 vector, bandwidth for bivariate density.
(If h2 <= 0 or not given, it is set to
standard deviation of (10/log(n))|(8*n^0.2))
|
| N | optional positive scalar, number of grid points for the
approximation of the nonparametric function.
(Default is 50)
|
| R | optional positive scalar, number of grid points for internal
approximations. (Default is 250)
|
| debug | optional scalar, should be set to 1 for plots during
computation. (Default is 0.)
|
Output: |
| npfit.g | N*N x 3 matrix, estimated function. |
| npfit.f | N*N x 3 matrix, estimated function of the corresponding ARMA model. |
| npfit.d1 | R x 2 matrix, estimated one-dimensional deconvolution
density estimator of the corresponding ARMA model. |
| npfit.d2 | N*N x 3 matrix, estimated two-dimensional combined
deconvolution/regular density estimator. |