pdh21/extreme_submm_sb_gals

Using xid+ with SED prior information to investigate possible extreme starburst galaxies identified from the blind Herschel SPIRE catalogues

extreme_submm_sb_gals's Issues

Use Papermill to pipeline the sources

PaperMill looks like the best way to run batches of notebooks for the Extreme Starburst Sources.

CIGALE emulator

First prototype, using log10 scale on sfr and redshift and fluxes (MIPS +SPIRE)

One source investigation

In order to see whats going on with adding more data like IRAC and MIPS, I could crate model map with One source

Identify sources to fit

I need to go back and look at what sources I was fitting originally.

SED fits.

Need code that links cigale so that I can produce SEDs of fits. Its going to require A LOT of seds to be produced (e.g. however many samples there are in the posterior) so should only be done on an object by object basis

Move prototype code to module code

At the moment, the model is in a notebook. I need to write this up as a module, along with the fitting and posterior code

suggestions from Seb

• Do zscore histograms at different sfr's
• See if Cigale can fit the opt/nir of my objects to get an idea of parameters

Comparison with XID+ fluxes original.

To make case for going to simultanous emualtro approach, I could show how the fluxes shrink when going to emulator model. Even better would be to do it in parameter space e.g. SFR (though problem with that is to justify not using other data)

Principled Workflow

I need to carry out full principled workflow:

1. Sample from prior (done)
2. Fit prior samples
3. Calibrate fits
4. z score and shrinkage

SIDES simulation tests

Zscores show a in issue at large SFR, with both z and sfr having negative scores

Test models

Having identified what galaxies to fit, I need to fit them with xid+

Analyse fits

I need to analyse fits, including:

• Contour plots of sources of interest
• SED plots
• Blind detection on replicated maps

Relook at potential candidates across all of HELP

I had created a plot looking at what blind sources had a bad fit. I could do for all of HELP