Setting the peak amplitude of a band

[elle-miller]

Hi there,

I am working on a project where I am combining two telescope surveys to analyse a SN event. This is because the space-based survey (Kepler - kst) is very good at recovering t0, and the other ground-based survey (Pan-Starrs - psg, psr, psi) is good at measuring colour c. It is my hope that by combining the differing strengths of these surveys, a more accurate fit can be obtained. I am using the SALT2 model (version 2.4), setting redshift and allowing x0, t0, c and x1 to vary.

For example, take the kst data and fit for SN2018oh.

Now, here is the Pan-Starrs data and fit for the same object.

However when I combine these data sets together, the fit for the data from Pan-Starrs (ps) isn't reaching the peak as it should. I am facing this issue with the majority of my sample.

I have tried various means to fix this, mostly by trying to force the Pan-Starrs data to a specified peak, but with not much success:

• Set guess_amplitude = False (slightly better fit but still not great)
• Binned the kst data down to only a couple of points so same 'weighting' as ps (no difference)
• Set peak magnitude in one band before the chi2 fitting process using model.set_source_peakabsmag (no change in output magnitudes)
• Set x1,c to values from Pan-Starrs only fit, then changed model magnitude in one band. I have found that this affects the other model magnitudes, but for some reason has no impact on the fitted model magnitudes

Do you have any ideas about what could be the problem here? I am particularly interested in being able to tell the fitter a starting guess for the amplitude, but the methods I have tried seem to be doing nothing.

Any help would be greatly appreciated! Thanks in advance.

Elle

[kbarbary]

Hi Elle, Have you tried fitting t0 based on the Kepler data alone, then fit the Pan-Starrs data (without Kepler) with t0 fixed to the fitted value? My guess is that the Kepler data has such small error bars and so many points that it completely dominates the fit (the other bands contribute only negligibly to the total likelihood). So whenever Kepler is included, the fitter tries to satisfy it and disregards other bands. Even if you bin the data, the small error bars are still a problem. You could synthetically increase them, but you'd have to do so very carefully to get just the result you want. The root problem is that the Kepler data seems to disagree with the color law favored by the other bands. It could be that this SN has a weird color law but that seems unlikely because the other bands all agree well with the SALT2 color law. It could be that the SN template is poorly characterized in some region covered by the (I assume wide) Kepler bandpass. Or the Kepler bandpass is poorly measured. Or there's a standards calibration offset between Kepler and Pan-Starrs. The last is my best (not-very-informed!) guess. Best wishes, -- Kyle

…

On Sun, Jan 5, 2020, 4:41 PM elle-miller ***@***.***> wrote: Hi there, I am working on a project where I am combining two telescope surveys to analyse a SN event. This is because the space-based survey (Kepler - kst) is very good at recovering t0, and the other ground-based survey (Pan-Starrs - psg, psr, psi) is good at measuring colour c. It is my hope that by combining the differing strengths of these surveys, a more accurate fit can be obtained. I am using the SALT2 model (version 2.4), setting redshift and allowing x0, t0, c and x1 to vary. For example, take the kst data and fit for SN2018oh. [image: image] <https://user-images.githubusercontent.com/55350122/71787839-7ecc2580-3070-11ea-8b57-333e30f9111e.png> Now, here is the Pan-Starrs data and fit for the same object. [image: image] <https://user-images.githubusercontent.com/55350122/71788007-8c82aa80-3072-11ea-8585-00e7d5ab3183.png> However when I combine these data sets together, the fit for the data from Pan-Starrs (ps) isn't reaching the peak as it should. I am facing this issue with the majority of my sample. [image: image] <https://user-images.githubusercontent.com/55350122/71788040-e97e6080-3072-11ea-885f-34aa9aebb277.png> I have tried various means to fix this, mostly by trying to force the Pan-Starrs data to a specified peak, but with not much success: - Set guess_amplitude = False (slightly better fit but still not great) - Binned the kst data down to only a couple of points so same 'weighting' as ps (no difference) - Set peak magnitude in one band before the chi2 fitting process using *model.set_source_peakabsmag* (no change in output magnitudes) - Set x1,c to values from Pan-Starrs only fit, then changed model magnitude in one band. I have found that this affects the other model magnitudes, but for some reason has no impact on the fitted model magnitudes Do you have any ideas about what could be the problem here? I am particularly interested in being able to tell the fitter a starting guess for the amplitude, but the methods I have tried seem to be doing nothing. Any help would be greatly appreciated! Thanks in advance. Elle — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#254?email_source=notifications&email_token=AAHM3F3TDMMZLZFOTUHZGNDQ4J42FA5CNFSM4KC6BGEKYY3PNVWWK3TUL52HS4DFUVEXG43VMWVGG33NNVSW45C7NFSM4IEDQ2BA>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAHM3F6JTUGHQXJT2RTDF4DQ4J42FANCNFSM4KC6BGEA> .