What is the actual algorithm that Gretl uses to estimate the Hurst exponent?
Is it the Anis-Lloyd corrected R/S or the DFA?

On Thu, Mar 25, 2021 at 11:03 AM Periklis Gogas <perrygogas@gmail.com> wrote:
Yes, exactly. And by sample size here we do not mean the initial sample lets say the 2000 observations I have, but the log_2(2000) I guess, the number of bins created.

On Thu, Mar 25, 2021 at 10:04 AM Sven Schreiber <svetosch@gmx.net> wrote:
Am 23.03.2021 um 12:18 schrieb Periklis Gogas:
> I just saw this thanks! are these treated to create confidence intervals
> as usual? ie Hurst +/-1.96* standard error?

Well, I'd cautiously say it would be good if you as the researcher had
some idea about the limiting distributions involved there. I'm not an
expert in that literature, but my guess would be that typically, yes, it
would be valid as an asymptotic approximation and almost surely not
strictly valid in finite samples.

Gretl-users mailing list -- gretl-users@gretlml.univpm.it
To unsubscribe send an email to gretl-users-leave@gretlml.univpm.it
Website: https://gretlml.univpm.it/postorius/lists/gretl-users.gretlml.univpm.it/