[Gretl-devel] Re: ACF and documentation of Bartlett standard errors
On Thu, Oct 10, 2024 at 4:48 AM Sven Schreiber
<sven.schreiber(a)fu-berlin.de> wrote:
[about Bartlett standard errors in the context of the "corrgm" command
and the GUI ACF plot.]
OK, so no discussion took off. I think there are two sub-issues
here:
First, the apparently missing documentation for the basic ACF/PACF
functionality. (Unless I overlooked something.) Secondly, pointing out
to the user what exactly the Bartlett standard errors there mean. (Also
a documentation thing.)
Any comments? Since I guess we're not going to write the documentation
for the upcoming release, I'm inclined to create tickets for these
issues, to come back to them in the future.
Good idea; this does need attention. FWIW here's what statsmodels says
(statsmodels.tsa.stattools.acf):
<quote>
If the autocorrelations are being used to test for randomness of
residuals as part of the ARIMA routine, the standard errors are
determined assuming the residuals are white noise. The approximate
formula for any lag is that standard error of each r_k = 1/sqrt(N).
For the ACF of raw data, the standard error at a lag k is found as if
the right model was an MA(k-1). This allows the possible
interpretation that if all autocorrelations past a certain lag are
within the limits, the model might be an MA of order defined by the
last significant autocorrelation. In this case, a moving average model
is assumed for the data and the standard errors for the confidence
intervals should be generated using Bartlett's formula.
</quote>
Allin