Bibtex

[JulienSiems]

I am using your code and it would be good to able to cite your repo! Could you post how you'd like it to be cited?

Thanks for your great work!

[josemiotto]

Hi Julien!

You can cite this zenodo link, which has a doi, or the paper associated with it (there is a link on the zenodo page)

https://zenodo.org/record/53787#.Yb8KzEYo-Nx

I realize that it has an old version, I.e. not synced with github. When I come back from vacations in a week I will update it, if you want to wait.

Best, and thanks, José

On Thu, 16 Dec 2021, 17:57 Julien, @.***> wrote:

I am using your code and it would be good to able to cite your repo! Could you post how you'd like it to be cited?

Thanks for your great work!

— Reply to this email directly, view it on GitHub https://github.com/josemiotto/pylevy/issues/17, or unsubscribe https://github.com/notifications/unsubscribe-auth/AACCTUUHH2ARE6E72KHRAMDURILBBANCNFSM5KGZ2ASQ . Triage notifications on the go with GitHub Mobile for iOS https://apps.apple.com/app/apple-store/id1477376905?ct=notification-email&mt=8&pt=524675 or Android https://play.google.com/store/apps/details?id=com.github.android&referrer=utm_campaign%3Dnotification-email%26utm_medium%3Demail%26utm_source%3Dgithub.

You are receiving this because you are subscribed to this thread.Message ID: @.***>

[JulienSiems]

Hi José,

Great, no rush! Let me know once you have updated it. I am pretty surprised how well fitting a stable distribution via your tabulated values works. I compared against a more classical estimator for the tail index for samples of a stable distribution with different tail indices and skewness and your estimator was an order of magnitude more accurate in terms of mean absolute error.

I did notice some outliers in the estimation, though. It may help to implement retries for the optimization in scipy or use evolutionary algorithm, like differential evolution, rather than gradient-based optimization?

Best, Julien

[josemiotto]

Hi,

It is surprising indeed haha. I don't know what estimators you are using but the classic problem there is to choose a boundary for the tail.

Can I ask you for what are you using it?

Best, José

On Sun, 19 Dec 2021, 12:09 Julien, @.***> wrote:

Hi José,

Great, no rush! Let me know once you have updated it. I am pretty surprised how well fitting a stable distribution via your tabulated values works. I compared against a more classical estimator for the tail index for samples of a stable distribution with different tail indices and skewness and your estimator was an order of magnitude more accurate in terms of mean absolute error.

Best, Julien

— Reply to this email directly, view it on GitHub https://github.com/josemiotto/pylevy/issues/17#issuecomment-997372748, or unsubscribe https://github.com/notifications/unsubscribe-auth/AACCTUSTKGC5XMJN5QQCKCLURW4QBANCNFSM5KGZ2ASQ . Triage notifications on the go with GitHub Mobile for iOS https://apps.apple.com/app/apple-store/id1477376905?ct=notification-email&mt=8&pt=524675 or Android https://play.google.com/store/apps/details?id=com.github.android&referrer=utm_campaign%3Dnotification-email%26utm_medium%3Demail%26utm_source%3Dgithub.

You are receiving this because you commented.Message ID: @.***>

[JulienSiems]

Hi José,

but the classic problem there is to choose a boundary for the tail.

Do you mean to find the tail index by fitting an exponential decay to the tails?

I am using this one: https://link.springer.com/article/10.1007/s00184-014-0515-7 (Corollary 2.4) It's for example used quite extensively in: https://arxiv.org/abs/1901.06053 Below, is the plot I meant for the absolute alpha estimation error for different stable samples.

Best,

Julien