Wavelet decompositions for the gmwm2 package

[robertomolinari]

Hey everyone,

I'll let you guys (Justin and Nate) coordinate, but the tasks that should be carried out now (given that the package can be installed) are as follows:

• [x] test whether the brickwall function is needed or if we can just apply the filters up to a certain number of observations for each scale (T - L_j + 1) where T is the length of the time series and L_j is the length of the filter at the j^th scale.
• [x] create a function for the maximum overlap (modwt) using the same function from the gmwm package.
• [x] create functions that takes a multivariate time series as an argument, applies the dwt or the modwt functions to each individual time series (that can have different lengths from each other) and outputs the decomposition for each time series. We would probably have to input and output list-type arguments (given that the time series can be of different lengths) but you probably know a better approach.
• [ ] I've given access to the spgmwm repository where you can find the spatial wavelet decomposition (see the sp_wv_coeffs() function in the spatial_wv.cpp file). This gives the decomposition using a matrix of spatial data as an input (check the vignettes). It is important to understand if this decomposition can be made in a more efficient manner. Check how the filtering changes compared to the pyramid algorithm and see if it can be adapted to this case.

Remember to enter the information and description for each function/file you create in order for them to be well documented. Once these functions work we can start creating some template examples that show how to use them.

[munsheet]

Implemented so brickwall not needed for dwt, still working on case for modwt

[munsheet]

Need more information on reflection coefficients!

[stephaneguerrier]

@robertomolinari Could you please document the function in the file: "sp_modwt_rough.r". Thanks!