Vonmises moments

[Domenico89]

Add moments and tests for the below distributions as part of #5078:
-pymc.distributions.continuous.VonMises

[Domenico89]

The curren tests should work fine, but if I use np.arange(5) as mu in line 446 ,f or example, I get as a result array([ 0., 1., 2., 3., -2.28318531]), which is not what I would expect. I couldn't figure out where -2.2831853 is coming from. Other strange numbers appear when I run the test for np.arange(k) with k>4.

[ricardoV94]

The curren tests should work fine, but if I use np.arange(5) as mu in line 446 ,f or example, I get as a result array([ 0., 1., 2., 3., -2.28318531]), which is not what I would expect. I couldn't figure out where -2.2831853 is coming from. Other strange numbers appear when I run the test for np.arange(k) with k>4.

Seems like it is just mirroring around +-np.pi? Are you testing the moment with a transformed distribution?

[codecov]

Codecov Report

Merging #5232 (1d337ca) into main (a16ec4a) will decrease coverage by 0.00%.
The diff coverage is 100.00%.

@@            Coverage Diff             @@
##             main    #5232      +/-   ##
==========================================
- Coverage   78.95%   78.94%   -0.01%     
==========================================
  Files          88       88              
  Lines       14232    14237       +5     
==========================================
+ Hits        11237    11240       +3     
- Misses       2995     2997       +2     

Impacted Files	Coverage Δ
pymc/distributions/continuous.py	96.72% <100.00%> (+0.12%)	⬆️
pymc/parallel_sampling.py	86.33% <0.00%> (-1.00%)	⬇️

[ricardoV94]

/pre-commit-run

[ricardoV94]

All tests are passing except for a pre-commit issue. I added a commit fix it with the bot, so make sure to merge from here if you want to do anything with this branch.

I guess you were seeing the effects of the CircularTransform?

with pm.Model() as m:
  x = pm.Vonmises("x", np.arange(4))  
  y = pm.Vonmises("y", np.arange(4), transform=None)

The moment of x will by default be on the Circular scale, whereas that of y will not. We are testing with the second case, when we call this line:

pymc/pymc/tests/test_distributions_moments.py

Line 90 in 1d337ca

return_transformed=False,

[ricardoV94]

Thanks @Domenico89, everything seemed to be working in our tests here.

Let me know if you find issues that need to be fixed or if my hunch above was correct.

[Domenico89]

Thanks @Domenico89, everything seemed to be working in our tests here.

Let me know if you find issues that need to be fixed or if my hunch above was correct.

ah, ok. Sorry if took this long, I am new with the base code. Thanks for the hunch!