Uncertainties in data

[stefanocovino]

Dear friends,

I have read the paper where this algorithm was described some time ago but now I am only now trying to apply it for real data in my field. I’m an astrophysicist and I am evaluating the PSD of a long light curve for a quasar.

And my question is simple. I see that, at present at least, the analyzed time series cannot have errors, is this an intrinsic feature of the method or is it just because in many fields an assumption of homoscedasticity is reasonable?

Needless to say that I would be interested in adding the errors to the data points to the analysis.

Cheers,
Stefano

[felipe-tobar]

Dear Stefano, many thanks for your email. The issue that you point out regarding known variances of measurements is indeed of interest and has been indicated to me by my colleagues from astronomy as well. I am not fullt dedicated to that project, but I will be happy to collaborate. I would be interested on research (let's think perhaps on a short article) from which the code can follow rather than just code? In general, as far as I am concerned, Gaussian processes do not exploit the information about the measurement variance. Let me know if you are interested and have the time to work on this line of research. best, Felipe Tobar, Universidad de Chile

…

On 06-10-20 04:59, stefanocovino wrote: Dear friends, I have read the paper where this algorithm was described some time ago but now I am only now trying to apply it for real data in my field. I’m an astrophysicist and I am evaluating the PSD of a long light curve for a quasar. And my question is simple. I see that, at present at least, the analyzed time series cannot have errors, is this an intrinsic feature of the method or is it just because in many fields an assumption of homoscedasticity is reasonable? Needless to say that I would be interested in adding the errors to the data points to the analysis. Cheers, Stefano — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#1>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ACT3KGYNY5MUIW7HQP7CMJLSJLE6LANCNFSM4SFVOOYQ>.

[stefanocovino]

Dear Felix, please correct me if I am misunderstanding. Are you proposing to think to some astrophysical problem that could be proftable addressed following your idea? If this is so, I could not hope to a better proposal, and I gladly accept. Of course we are both following our projects, yet I think it will not be difficult to find some time for it. At present, I am interested in methods to address the significance of possible QPOs in irregulary sampled time series affected by correlated noise, but there are for sure other scenarios where deriving a non-parametric PSD can help. We can discuss further, of course. Chers, Stefano P.S. Indeed I used to visit Chile for my research very often in the past, unfortunately this is somehting one has to postpone for a while now. Il mer 7 ott 2020, 19:59 felipe-tobar <notifications@github.com> ha scritto:

…

Dear Stefano, many thanks for your email. The issue that you point out regarding known variances of measurements is indeed of interest and has been indicated to me by my colleagues from astronomy as well. I am not fullt dedicated to that project, but I will be happy to collaborate. I would be interested on research (let's think perhaps on a short article) from which the code can follow rather than just code? In general, as far as I am concerned, Gaussian processes do not exploit the information about the measurement variance. Let me know if you are interested and have the time to work on this line of research. best, Felipe Tobar, Universidad de Chile On 06-10-20 04:59, stefanocovino wrote: > > Dear friends, > > I have read the paper where this algorithm was described some time ago > but now I am only now trying to apply it for real data in my field. > I’m an astrophysicist and I am evaluating the PSD of a long light > curve for a quasar. > > And my question is simple. I see that, at present at least, the > analyzed time series cannot have errors, is this an intrinsic feature > of the method or is it just because in many fields an assumption of > homoscedasticity is reasonable? > > Needless to say that I would be interested in adding the errors to the > data points to the analysis. > > Cheers, > Stefano > > — > You are receiving this because you are subscribed to this thread. > Reply to this email directly, view it on GitHub > <#1>, > or unsubscribe > < https://github.com/notifications/unsubscribe-auth/ACT3KGYNY5MUIW7HQP7CMJLSJLE6LANCNFSM4SFVOOYQ >. > — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub <#1 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAGVYHCBALAL25XWVLAQHATSJSUBPANCNFSM4SFVOOYQ> .