Re: [Gretl-devel] Getting ACF/PACF/XCF as a matrix
On Mon, 2 Feb 2009, Riccardo (Jack) Lucchetti wrote:
On Mon, 2 Feb 2009, Sven Schreiber wrote:
> On 02.02.2009 20:55, Riccardo (Jack) Lucchetti wrote:
>>
>> We could actually drop the extra boolean parameter by using a
>> trick similar to what we currently do for ADF: p>0 means
>> autocorrelations, p<0 means partial autocorrelations. Or
>> would it be too arcane?
>
> yes, FWIW I wouldn't like that very much
It's no big deal on my side; it was mainly an aesthetic thing: I
don't like very much having a string parameter dangling at the
end. Also, the positive/negative thing may simplify script
writing.
I think that aesthetics are pretty much in the eye of the beholder
here. Like Jack, I don't much like the trailing string parameter,
but neither do I much like flipping the sign of a parameter, if it
can be avoided. IMO the cleanest solution is a distinct function
name, but that has its drawbacks too. Urgh.
On overloaded functions, I'll just say, think about writing the
help entry for the function: if it seems too crazy, give up the
idea of overloading.
> I don't think case 2 is so unlikely, but if the ACF is
> computed anyway then sure you can spit it out as well
Yeah, in practice we would have the options "just acf"/"pacf
too"
As Jack says, computing the PACF, as we do it, requires computing
the ACF first. However, unless this sort of calculation is
embedded in a loop with millions of iterations, I don't see CPU
time as a compelling consideration. Are users likely to want PACF
without ACF? I don't know -- but if they might, then I'd say let
them have it (and not be obliged to extract the second column of a
2-column matrix).
On the periodogram/smoothed spectrum issue, I'm agnostic, and will
simply wait till the dust settles ;-)
Allin.