search

SMAC-Group / gmwm

Generalized Method of Wavelet Moments (GMWM) is an estimation technique for the parameters of time series models. It uses the wavelet variance in a moment matching approach that makes it particularly suitable for the estimation of certain state-space models.
Other
28 stars 14 forks source link

Extend WV Filters #140

Closed coatless closed 8 years ago

coatless commented 8 years ago

Feature:

It would be very nice to provide support for MODWT and DWT based decomposition using wavelet filters other than Haar. (We are still restricting the GMWM to using the Haar decomposition.) With this change, we can effectively compute the WV under these different filters as well.

Note: Robustness will only work with the Haar filter.

Per 9af93bd4bcce465e478227bb5702caf5cb232323, we now have a very dynamic way of calling WV filters.

All that is needed is a translation of the filter coefficients into C++.

robertomolinari commented 8 years ago

Just to clear : we can’t use the GMWM with other filters until we have the corresponding theoretical form of WV (unless we use indirect inference). Otherwise, for simple WV computations I agree! Moreover, robustness works with other filters as well (we just require the relative wavelet coefficients to be stationary). If I misunderstood your email: sorry!!

coatless commented 8 years ago

There is no change to the GMWM procedure as we only have the Haar WV theoretical process forms loaded.

The motivation behind the WV filter changes is to expand decomposition options given by modwt(), dwt(), and wvar() (e.g. simple WV computations). As a result, the feature puts the gmwm into more of a platform role for wavelet research since it has more flexibility in the aforementioned functions.

robertomolinari commented 8 years ago

Ok, that’s fine! The robust theory for the WV will hold also in this case.