How to implement multiple 'observations'

[amn3142]

I.e. we expect the same target flux in different exposures but maybe a different background.
I believe this was a new feature that was added but sadly slack has eaten the discussion.

[sibirrer]

@amn3142, there is a parameter called flux_scaling

lenstronomy/lenstronomy/Data/imaging_data.py

Line 89 in 67949f0

:param flux_scaling: scales the model amplitudes to match the imaging data units. This can be used, for example,

such that you can effectively scale the fluxes from one band to another such that the same parameters have a different meaning on the flux count for different bands. You can then combine that with the joint-linear mode of simultaneously model multiple bands with the same amplitudes and they then get re-scaled according to that parameter (it is a multiplicative factor).

[amn3142]

Thanks very much! On Aug 27, 2023, at 8:01 AM, Simon Birrer ***@***.***> wrote: @amn3142<https://github.com/amn3142>, there is a parameter called flux_scaling https://github.com/lenstronomy/lenstronomy/blob/67949f01fa26f70aa551fef9fb853ce4c0baef20/lenstronomy/Data/imaging_data.py#L89 such that you can effectively scale the fluxes from one band to another such that the same parameters have a different meaning on the flux count for different bands. You can then combine that with the joint-linear mode of simultaneously model multiple bands with the same amplitudes and they then get re-scaled according to that parameter (it is a multiplicative factor). — Reply to this email directly, view it on GitHub<#510 (comment)>, or unsubscribe<https://github.com/notifications/unsubscribe-auth/AEA6XOUCNTTBKM57CX3B7ELXXNOLBANCNFSM6AAAAAA37D2HIA>. You are receiving this because you were mentioned.Message ID: ***@***.***>

…

-------------------------------------------- Anna M. Nierenberg Assistant Professor Department of Physics, UC Merced annanierenberg.com

[amn3142]

Ok right so the one limitation was that the sky background cannot be different by an additive factor right? So I will just need to fit and subtract that first. On Aug 27, 2023, at 10:44 AM, Anna Nierenberg ***@***.***> wrote: Thanks very much! On Aug 27, 2023, at 8:01 AM, Simon Birrer ***@***.***> wrote: @amn3142<https://github.com/amn3142>, there is a parameter called flux_scaling https://github.com/lenstronomy/lenstronomy/blob/67949f01fa26f70aa551fef9fb853ce4c0baef20/lenstronomy/Data/imaging_data.py#L89 such that you can effectively scale the fluxes from one band to another such that the same parameters have a different meaning on the flux count for different bands. You can then combine that with the joint-linear mode of simultaneously model multiple bands with the same amplitudes and they then get re-scaled according to that parameter (it is a multiplicative factor). — Reply to this email directly, view it on GitHub<#510 (comment)>, or unsubscribe<https://github.com/notifications/unsubscribe-auth/AEA6XOUCNTTBKM57CX3B7ELXXNOLBANCNFSM6AAAAAA37D2HIA>. You are receiving this because you were mentioned.Message ID: ***@***.***>

…

-------------------------------------------- Anna M. Nierenberg Assistant Professor Department of Physics, UC Merced annanierenberg.com

-------------------------------------------- Anna M. Nierenberg Assistant Professor Department of Physics, UC Merced annanierenberg.com

[sibirrer]

no, sky background and all other noise properties (incl. PSF) can be different from image/exposure to image/exposure anyways. The multiplicative factor is effectively compensating for the model predicted flux from a given surface brightness (effectively magnitude zero point differences)

[amn3142]

oh sorry to be clear, I meant the ‘uniform’ lens model parameter, not the noise. On Aug 27, 2023, at 10:48 AM, Simon Birrer ***@***.***> wrote: no, sky background and all other noise properties (incl. PSF) can be different from image/exposure to image/exposure anyways. The multiplicative factor is effectively compensating for the model predicted flux from a given surface brightness (effectively magnitude zero point differences) — Reply to this email directly, view it on GitHub<#510 (reply in thread)>, or unsubscribe<https://github.com/notifications/unsubscribe-auth/AEA6XOTANEITZAFRYLDFO73XXOCAPANCNFSM6AAAAAA37D2HIA>. You are receiving this because you were mentioned.Message ID: ***@***.***>

…

-------------------------------------------- Anna M. Nierenberg Assistant Professor Department of Physics, UC Merced annanierenberg.com

[sibirrer]

ah I see. Yes, you need that to be subtracted first as the current modeling only allows all of them being the same or none.

[amn3142]

Sorry one more question, is this flux scaling something that lenstronomy can fit for during the linear optimization? Or do I need to figure it out ahead of time?

[sibirrer]

you have to know it ahead of time (or perhaps we can enable it to be fit as a non-linear parameter) as the linear response vector across bands needs to be known before linear optimization can be used

[amn3142]

That would be great if we could fit for it directly during the linear optimization, we could also include a background parameter which would solve all of my problems.