Sun, 06/20/2021 - 12:25
I am completing some data analysis on a variable star and have good results finding the period.
However, can I get a period uncertainty (+/- ...days) in V-Star?
Any help is greatly appreciated.
That's not currently implemented in VStar. The DCDFT implementation is a port of Fortran code, which as far as I'm aware, does not provide this information either.
Other than drawing some conclusions based upon the resolution specified in the DCDFT parameters dialog or using the Current Mode ANOVA plugin (or use the AoV period search plugin and compare results with DCDFT), there is no intrinsic uncertainty information available.
A similar question has been raised in the past about polynomial fit in this forum. Right now, AIC, BIC, RMS and ANOVA can be applied for polyfit but again there's no intrinsic uncertainty information available.
I would welcome input from you or others in AAVSO re: the best way(s) to implement uncertainty analysis this for DCDFT and polynomial fit. I don't claim to have particular expertise in that area but I'm willing to learn. Brad Walter, for example, may well have specific thoughts about this.
Thank you and apologies for the slow reply. i just realised I hadn't subscribed to this thread.
I'll run the data through the different methods and see what I find.
No worries John. Happy to talk more about this.
A long time coming, but version 2.22.0 adds uncertainty measures for DCDFT derived Fourier models.
See p 67 and 68 of the user manual re: this.
There is a bug found since release that sometimes causes such models to fail:
Being worked on.
The manual that you access from the help menu of 2.22.0 is the manual for 2.21.0 and doesn't have info about uncertainty estimates for the DCDFT Models. However, if youuse the VStar online link in the help menu and click on the USER MANUAL bold heading and download the 10 mb manual using the download button on that github page you get the 2.22.0 manual which has info on the uncertainty estimates. Page 69 of the manual contains the following statement:
"In most cases the possible range estimated by the theoretical formula is too liberal while
the range estimated by FWHM is too conservative."
It is important to understand that "too liberal" means underestimates the error and "too conservative" means over estimates the error. By calculating both estimates one can get lower and upper bounds for the uncertainty.
Brad Walter, WBY
Thanks for that Brad. I added https://github.com/AAVSO/VStar/issues/298
We may be able to change the file being pointed to in the meantime.
And yes, implementing it that way was absolutely intentional re: upper/lower band, as per Foster.
Hello Brad & David,
Thanks for pointing that out. I just updated the copy of the User Manual that is linked from VStar's "Help" menu item so it should be OK now.