Re: [Gretl-users] Mixed Data Sampling Regression

Am 31.03.2016 um 16:24 schrieb Riccardo (Jack) Lucchetti: > On Thu, 31 Mar 2016, Jan Tille wrote: > please allow me a short question. Does anyone know whether there is a >> function package for Gretl that can perform MIDAS regression analysis, >> in the spirit of Eric Ghysels's Matlab toolbox or as provided in >> EViews 9.5? >> > What a coincidence. I was thinking of writing one just this morning. I'm > very busy with teaching atm, though, and there are a few other things on > the gretl fromnt that I'm working on now, so don't hold your breath :) > > Actually (I just re-checked), there is a big difference between Eviews and gretl in that respect: Eviews has workfiles with mixed frequencies (the frequency is a per-"page" property). Gretl only supports a single frequency in a workfile. Knowing Jack, he probably wants to do a lot of "join"-ing with immediate transformation to matrices. But I think that would not be good, because it would go against the user interface. Perhaps one way to implement mixed frequency workfiles would be to have a standard gretl workfile with the highest frequency (say, monthly), and then add lower-frequency series (say, quarterly) to that workfile either by padding with missings to get "gappy" series or by repeating the values, PLUS by introducing a new series property flag denoting the actual frequency. But this is not thought through.