add support for time dependent (and filter dependent) systematic error

[sahiljhawar]

This PR adds support for sampled systematic error and time (and filter) dependent (interpolated) systematic error. For optimum performance it is recommended that the filters to be used for systematic biases should be less than or equal to the no of filters being used for the inference. If the no of filters for systematic biases is more than the no of --filters then the sampler will unnecessarily sample the priors which are of no use. For eg. if the filters in --filters is u,g,r,i,z,y,J,H,K and you want to independently sample r and i filters then the yaml should be

config:
  withTime:
    value: true
    filters: ["r", "i", null]
    ...
    ...

null will be a single array which will be used for all filter excluding r and i.

[sahiljhawar]

@mcoughlin @tsunhopang Can you please have a look at the ALLOWED_FILTERS in systematicsprior.py?

Right now I am using Peter's svd files and filters --filters u,g,r,i,z,y,J,H,K

[mcoughlin]

@sahiljhawar Is the issue that we don't have a list of the now standard filters somewhere?

[sahiljhawar]

@mcoughlin Yes, but apart from that. I am not sure which filters should be in the allowed list. Since sdss::u and uvot::u are different u bands.

[mcoughlin]

I thought we were trying to no longer allow folks to use bands without their full names?

[sahiljhawar]

Okay, in that case how can filters be named as if they are variable names and not strings. Is there any standard replacements for the symbols? This question includes the Xrays, etc filters as well.

[mcoughlin]

Not sure I understand the question, but we do have code here: https://github.com/nuclear-multimessenger-astronomy/nmma/blob/main/nmma/em/utils.py#L153

[sahiljhawar]

Basically a list of filters which is pass to —-filters

something like (filter names) which is suffixed to the models on zenodo.

[mcoughlin]

Maybe we just hard code it in list form somewhere then

[sahiljhawar]

Seems like a good idea.