Utils tests

[jeffjennings]

Adds tests for the routines in utilities.py that aren't covered by existing tests. These tests check that the routines simply run and that they give the expected numerical results.

Closes #214

[jeffjennings]

The other program that has low test coverage is filter.py; we should probably add tests for those missing routines (three of the functions in CriticalFilter). Do you want to do that (unless you expect those functions to change)? Or I can.

[rbooth200]

I can add regression tests for two of the functions in filter.py. The function update_power_spectrum_multiple isn't currently used anywhere so I'd leave it untested. Can you check the maths for the other two? The covariance function should be the same as the Hessian in the frank debris paper and the prior is log P(p|beta) in the frank paper.

[jeffjennings]

I can add regression tests for two of the functions in filter.py. The function update_power_spectrum_multiple isn't currently used anywhere so I'd leave it untested. Can you check the maths for the other two? The covariance function should be the same as the Hessian in the frank debris paper and the prior is log P(p|beta) in the frank paper.

Great, changes to filter.py should probably be a separate PR.

Re: math: for log_prior, in the docstring should the - np.sum(p0/pi + (alpha-1)*np.log(p0/pi)) be (change to the last log term) - np.sum(p0/pi + (alpha-1)*np.log(pi) -- (eqns.37-40 in the paper)? If it's correct as it is, the docstring and code are consistent.

For covariance_MAP, (eqn.40 in the debris paper), I'm confused where the 2 comes from in (2 * mqq + Dqq) in l.218, - 0.5 * np.outer(1 / p, 1 / p) * (2 * mqq + Dqq) * Dqq, but otherwise it looks right.

[rbooth200]

I checked that term - it's the first line in eqn 30, so I think everything is fine. I've added a test that checks both of the missing functions in filter.py and some of the other likelihood functions. We're now up to 86% coverage with ~half the missing lines coming from stats models and make_figs

[rbooth200]

Approving this. If you're happy with the filter tests, feel free to merge.

[jeffjennings]

Great, thank you