Possible Flux Offset - probably due to Peculiar Velocity. Sims take mu as arg?

[RobFirth]

Flux offset between real and simulated data, for example SN2009jf here:

Is this a result of scaling to -17.0? I thought we decided against scaling in the end?

[RobFirth]

and here with an offset of +2.9 mag (!):

[sprajs]

After digging though most of the code structure it seems like the problem is related to MPFIT in some way. Light curve fitting works exactly as expected when performed with MultiNest. This is, however, computationally prohibitive for a large simulations so we will need to get to the bottom of the problem. MPFIT fails to perform a fit entirely which is why the light curve looks like a SN but doesn't match the observations.

[sprajs]

Stack overflow ... the error, not the website! The number of data points wasn't reported correctly to MPFIT which meant that it was reading the array of residuals as being whatever was the next 100 data point in memory were.

[sprajs]

That fixed one issue but the problem still remains, it now looks like the fit parameters do not get updated after each pass off MPFIT. This will inevitably end up being a pointer dereference issue as per usual.

[RobFirth]

https://seal.web.cern.ch/seal/snapshot/work-packages/mathlibs/minuit/

[sprajs]

ROOT::Minuit2::Simplex behaves in exactly the same way as MPFIT. Thankfully there are a few more types of minimisers that we can try in Minuit2

[sprajs]

The flux offsets issue in now solved in the latest develop branch. The fix was to replace MPFIT with ROOT::Minuit2::Minos. Light curves are still not exactly right but the root of that issue is in issue #5 .

[RobFirth]

Looks like this issue might be back.

SN1994I

---

Photometry

Photometry looks like:

Specphot

The specphot, generated using

flux, flux_err = coco.simulate(b"SN1994I", 
                    z_obs, 0.0, 0.0, 0.0, 3.1, 
                    mjdmax, mjd_to_sim, 
                    filters_to_sim)

looks like:

note y axis. Offset of a factor of ~2.

For reference, phot with spectra and fits from LCFIT

[RobFirth]

On the plus side, colour looks bob-on

[RobFirth]

So this is probably an issue with a mismatch between redshift, z and distance modulus. It would be good for testing for the sim functions to accept the distmod directly.