Cover more priors with empirical distribution extension function

[AaronDJohnson]

This adds support for Normal and Uniform priors to be covered automatically by empirical distributions. Additionally, I have added a skip if the empirical distribution is outside of these functions with a warning that asks the user to make sure they are covering their priors.

Priors with a size attribute should also be covered here (e.g. free spectral red noise models).

[AaronDJohnson]

Hopefully fixing #179

[codecov-commenter]

Codecov Report

Merging #180 (ca633f4) into master (b7b90e5) will decrease coverage by 0.09%.
The diff coverage is 45.83%.

@@            Coverage Diff             @@
##           master     #180      +/-   ##
==========================================
- Coverage   37.06%   36.96%   -0.10%     
==========================================
  Files          20       20              
  Lines        3899     3931      +32     
==========================================
+ Hits         1445     1453       +8     
- Misses       2454     2478      +24     

Impacted Files	Coverage Δ
enterprise_extensions/empirical_distr.py	73.86% <ø> (ø)
enterprise_extensions/sampler.py	38.40% <45.83%> (-0.64%)	⬇️
enterprise_extensions/deterministic.py	11.83% <0.00%> (ø)

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update b7b90e5...ca633f4. Read the comment docs.

[Hazboun6]

We need to fix the handling of array-sampled parameters, specifically if there are single-sampled parameters later in the param list.

[Hazboun6]

I am testing this with a run that has an array-sampled parameter and I'm get an error at this line. I believe it's the same issue that the full param_names is longer than the list params. In my case there is a single-sampled parameter in the list after the array-sampled parameter. I think that would cause this problem. This is why in #179 I made a separate list of param names from the parameter.Parameter.name attributes.

The Trace of my error is below, if useful.

Traceback (most recent call last):
  File "/mmfs1/home/hazboun/miniconda3/envs/pta/lib/python3.9/site-packages/pta_sim/scripts//model2a_advnoise_sw.py", line 196, in <module>
    Sampler = sampler.setup_sampler(pta_crn, outdir=args.outdir, resume=True,
  File "/mmfs1/home/hazboun/miniconda3/envs/pta/lib/python3.9/site-packages/enterprise_extensions/sampler.py", line 1107, in setup_sampler
    jp = JumpProposal(pta, empirical_distr=empirical_distr, save_ext_dists=save_ext_dists, outdir=outdir)
  File "/mmfs1/home/hazboun/miniconda3/envs/pta/lib/python3.9/site-packages/enterprise_extensions/sampler.py", line 252, in __init__
    self.empirical_distr = extend_emp_dists(pta, self.empirical_distr, npoints=100_000,
  File "/mmfs1/home/hazboun/miniconda3/envs/pta/lib/python3.9/site-packages/enterprise_extensions/sampler.py", line 97, in extend_emp_dists
    if pta.params[param_idx].type not in ['uniform', 'normal']:
IndexError: list index out of range

[Hazboun6]

Also, this PR probably needs a unit test. They could be tacked onto the other empirical distribution tests.

[Hazboun6]

Alright @AaronDJohnson getting closer! It looks like here is just a typo throwing things for a loop. I'll try and get a simple case set up for testing. Sorry I haven't yet.

Traceback (most recent call last):
  File "/mmfs1/home/hazboun/miniconda3/envs/pta/lib/python3.9/site-packages/pta_sim/scripts//model2a_advnoise_sw.py", line 196, in <module>
    Sampler = sampler.setup_sampler(pta_crn, outdir=args.outdir, resume=True,
  File "/mmfs1/home/hazboun/miniconda3/envs/pta/lib/python3.9/site-packages/enterprise_extensions/sampler.py", line 1113, in setup_sampler
    jp = JumpProposal(pta, empirical_distr=empirical_distr, save_ext_dists=save_ext_dists, outdir=outdir)
  File "/mmfs1/home/hazboun/miniconda3/envs/pta/lib/python3.9/site-packages/enterprise_extensions/sampler.py", line 258, in __init__
    self.empirical_distr = extend_emp_dists(pta, self.empirical_distr, npoints=100_000,
  File "/mmfs1/home/hazboun/miniconda3/envs/pta/lib/python3.9/site-packages/enterprise_extensions/sampler.py", line 97, in extend_emp_dists
    short_par = '_'.join(emp_dist.param_namem.split('_')[:-1])  # make sure we aren't skipping priors with size!=None
AttributeError: 'EmpiricalDistribution1D' object has no attribute 'param_namem'
Traceback (most recent call last):

[AaronDJohnson]

Oops! I thought it was merged already. I'll undo the delete and reopen the PR 👍