improved modeling of galaxies with broad+narrow line-emission

[moustakas]

[WIP]

Algorithmic improvements in order to address #118. Details to follow.

[coveralls]

Pull Request Test Coverage Report for Build 5738059619

Warning: This coverage report may be inaccurate.

This pull request's base commit is no longer the HEAD commit of its target branch. This means it includes changes from outside the original pull request, including, potentially, unrelated coverage changes.

• For more information on this, see Tracking coverage changes with pull request builds.
• To avoid this issue with future PRs, see these Recommended CI Configurations.
• For a quick fix, rebase this PR at GitHub. Your next report should be accurate.

Details

• 77 of 105 (73.33%) changed or added relevant lines in 3 files are covered.
• 5 unchanged lines in 1 file lost coverage.
• Overall coverage increased (+0.1%) to 51.752%

---

Changes Missing Coverage	Covered Lines	Changed/Added Lines	%
py/fastspecfit/fastspecfit.py	0	2	0.0%
py/fastspecfit/emlines.py	71	97	73.2%
<!--	Total:	77	105	73.33%	-->

Files with Coverage Reduction	New Missed Lines	%
py/fastspecfit/emlines.py	5	51.03%
<!--	Total:	5		-->

Totals
Change from base Build 5719900769:	0.1%
Covered Lines:	2378
Relevant Lines:	4595

---

💛 - Coveralls

[moustakas]

This PR implements some key improvements in the broad-line fitting based on conversations with @stephjuneau, @Ragadeepika-Pucha, @paulmartini, and others. The key developments / insights are:

• DELTA_LINECHI2 records the raw (not reduced) delta-chi2 (DELTA_LINERCHI2 is now deprecated), given by the chi2 without broad lines minus the chi2 with broad lines.
• DELTA_LINENDOF (new to the data model) records the difference in the number of degrees of freedom between the model choices without and with broad lines, respectively.
• Both DELTA_LINECHI2 and DELTA_LINENDOF are measured from pixels around the Balmer and helium lines (rather than over the full spectrum).
• The narrow+broad-line model is now chosen if all of the following criteria are met:
  • DELTA_LINECHI2>DELTA_LINENDOF (i.e., the chi2 improvement is larger than the difference in the number of degrees of freedom between the two models);
  • SIGMA_BROAD > 250 km/s; and
  • SIGMA_BROAD > SIGMA_NARROW.

The last two criteria are needed to prevent broad lines from over-fitting lines in (especially) low-mass dwarf galaxies or to capture unmodeled residuals in the stellar continuum.

Note that there is no longer any formal S/N cut on the broad-line amplitude. (Previously the two reddest broad Balmer lines, e.g., H-alpha and H-beta for lower-redshift objects, were required to be detected with >3-sigma significance, which threw out many bona fide Seyfert 1.5s highlighted in #118.)

Here's a quick comparison of the statistics for a single healpix, sv3/bright/9428 (where 2.1.2 is the version of the code used to build the fuji/v2.0 VAC):