Re: [Gretl-users] Periodogram Values - Comparison with Statistica 9.0
On Fri, 5 Feb 2010, Riccardo (Jack) Lucchetti wrote:
On Fri, 5 Feb 2010, savos schmagges wrote:
> Hello
>
> i have a question concerning the periodogram values calculated by gretl.
> I have checked the values with Statistica, but they are differing by the factor 12.
> What is the reason for that? How can the values be reconciled?
I suspect it's a different normalisation convention. Try this:
<script>
nulldata 100
setobs 1 1 --special
x = fracdiff(normal(), -0.3)
a = pergm(x)
F = fft({x})
F = a ~ sumr(F[2:51].^2)/50
print F
</script>
and see if the values in the third column of F agree with those from
Statistica. You may want to take a look at section 5.10 of the Gretl
User's Guide.
One other note: William Greene has a periodogram example in the
chapter on Times Series Models in his Econometric Analysis (ch. 20
in the 5th edition). Gretl replicates that example (i.e. the
spectrum values are on the same scale), so one can look at his
text for a bit more detail on the conventions used.
Allin Cottrell