This is a call for help: I've finally managed to package in a semi-usable state the functions I have for multivariate spectra of time series. I have _not_ put the package onto the official repository yet because I don't think it's ready for production use. For example, documentation is in a rather dismal state. It would be very very nice if time-series savvy people could install the package and try things out. In order to do this, open a console and type install http://www2.econ.univpm.it/servizi/hpp/lucchetti/gretl/ghosts.zip and the "ghosts" package will appear in the list of available packages. Any kind of feedback is greatly appreciated. Thanks!!! ------------------------------------------------------- Riccardo (Jack) Lucchetti Dipartimento di Scienze Economiche e Sociali (DiSES) Università Politecnica delle Marche (formerly known as Università di Ancona) r.lucchetti(a)univpm.it http://www2.econ.univpm.it/servizi/hpp/lucchetti -------------------------------------------------------

Am 16.11.2017 um 11:29 schrieb Riccardo (Jack) Lucchetti: > > This is a call for help: I've finally managed to package in a semi-usable > state the functions I have for multivariate spectra of time series. I have > _not_ put the package onto the official repository yet because I don't > think it's ready for production use. For example, documentation is in a > rather dismal state. > Any kind of feedback is greatly appreciated. Thanks!!! I've briefly looked at it, and of course it's very nice to have that. (Before you had started on that I might have asked why not leave the background engine to R in this case, but since the work now basically seems to be done, there's no reason to toss it of course.) I don't have any systematic or fundamental comments, just some arbitrary reactions:

- csum(): looks trivial, why is it needed?

- It's not clear to me whether it's possible to get the Co- and quadrature spectra. I know that the main usage is to look at the coherence and phase, but still it might be useful in some cases. And the gain? (I guess it would be just a matter of writing more columns to the result matrix in co_ph().)

Am 29.11.2017 um 22:28 schrieb Riccardo (Jack) Lucchetti: > On Sun, 26 Nov 2017, Sven Schreiber wrote: >> - It's not clear to me whether it's possible to get the Co- and quadrature >> spectra. I know that the main usage is to look at the coherence and phase, >> but still it might be useful in some cases. And the gain? (I guess it >> would be just a matter of writing more columns to the result matrix in >> co_ph().) > > Well, I guess that in the very unlikely case you want to use those, you > just grab the "spectrum" element of the bundle and you apply the Re() and > Im() functions, respectively. Do you think this should go into the doc > file? I don't know what the "spectrum" element is or that is exists, so yes. In this pre-version of the help I just read about the "mvspec" element.