DOC: added examples to docstring of dirichlet class

[jeannefukumaru]

No description provided.

[jeannefukumaru]

This is my first pull request to an open-source project. Any pointers on how to improve are much appreciated!

[ilayn]

Hi @jeannefukumaru Thanks a lot for taking the time to prepare this. You are doing quite good so far. A few details to iron out:

• In the docs, numpy is implicit to all examples, you can skip import numpy as np line
• You might want to regroup similar items and decide on whether the explanations should be text or inline comment. Take ndimage.convolve as a random example. There are parts as explanations and parts as code (you can look at its source code how to do that). I think we can be a bit more economical here.
• you can skip the assignments to d just to print. Hence instead of

  >>> d = dirichlet.pdf(quantiles, alpha)
  >>> d  
  0.2843831684937255

  you can directly type

  >>> dirichlet.pdf(quantiles, alpha)
  0.2843831684937255

• I'll leave it up to your decision but either make it text or inline code comment for the initial 3 lines. Note that you can also use inline comments if it is short enough

  >>> dirichlet.rvs(alpha, size=2)  # get random samples from dirichlet
  array([[3.13387474e-02, 1.07631504e-01, 8.61029749e-01],
     [5.28822173e-05, 3.21909147e-01, 6.78037971e-01]])

[ilayn]

You can also track what went wrong in the CI suite, for this case these are the documentation errors detected

https://travis-ci.org/scipy/scipy/jobs/436477657#L1369

[ilayn]

Thanks @jeannefukumaru, almost there, two three minor items remaining :)

• Currently it renders like this https://9221-1460385-gh.circle-artifacts.com/0/html-scipyorg/generated/scipy.stats.dirichlet.html#scipy.stats.dirichlet

that is because it needs empty lines also after the text blocks in the examples.

• And also this line

scipy/scipy/stats/_multivariate.py

Line 1385 in e780726

once we specify the dirichlet distribution we can then calculate quantities of interest

is too long for 79 character limit.

• The last one about PEP8-E261 which we don't actually track but since we are at it why not?

Otherwise it is good to go.

[jeannefukumaru]

Thanks for the detailed feedback! It was very helpful. I've committed the edits you suggested

[ilayn]

Due to the doctest requirements we have to add #may vary to places where the output can change. I've added those and removed the variable assignment to make things print.

[jeannefukumaru]

Apologies for the delay and thanks for your patience. I've added the random_state keyword to dirichlet.rvs.

After looking at the details for why the CI builds are failing, it seems like python 2.7 and python 3.6 are two environments that are throwing the errors. Looks like the git checkout -qf FETCH_HEAD is giving a service error

[ilayn]

Travis Python 3.6 failure seems like a real one. The expected value for size = 2 is wrong. I get the same value as Travis CI reports

>>> dirichlet.rvs(alpha, size=1, random_state=1)
array([[0.00766178, 0.24670518, 0.74563305]])

>>> dirichlet.rvs(alpha, size=2, random_state=2)
array([[0.01639427, 0.1292273 , 0.85437844],
       [0.00156917, 0.19033695, 0.80809388]])

[ilayn]

Since this is almost done, I've rebased it. If CI is happy it can go in.

[ilayn]

Barely made it. Thanks @jeannefukumaru

[jeannefukumaru]

Thanks so much for your patience @ilayn :)