On 03.02.2009 03:00, Allin Cottrell wrote:
On the periodogram/smoothed spectrum issue, I'm agnostic, and
simply wait till the dust settles ;-)
Perhaps I should explain a little. There are (at least) two unrelated
First, it is my understanding that a design principle of gretl is that
every command should be accessible via GUI as well as via CLI. So it
struck me as awkward that you can get the smoothed spectrum via GUI but
not via CLI.
Second, what I'm interested in in the medium term are other
classic/standard/whatever-you-want-to-call-it frequency-domain measures
like coherency, phase, gain, and also multivariate generalizations like
multiple and partial coherencies etc. For the bivariate stuff like
simple coherency I have exploited gretl's new R-interface and used R.
The multivariate things need to be programmed, and for that you need the
estimators of the spectrum (and cross spectrum), not the periodogram AFAIK.
Of course there is a problem in that the cross spectrum is
complex-valued and I'm not sure at this point whether it is doable in
gretl without too many awkward workarounds. The obvious alternative is
to program in R, but I'm still far from fluent in R.