DOC: Slightly more clarification about the expected target value indexing of pymc.OrderedLogistic

[Mithrillion]

Issue with current documentation:

While it is possible to deduce that OrderedLogistic expects the observed values to have the typical 0-based indexing as in regular Python, based on the documentation's example:
y = np.concatenate((1*np.ones(n1_c), 2*np.ones(n2_c), 3*np.ones(n3_c))) - 1
the initial line of the documentation Useful for regression on ordinal data values whose values range from 1 to K as a function of some predictor... makes it seem like the encoding should be "1,2,3,...K". This can be confusing for users, as this can result in logp=-inf error, which does not intuitively trace back to the source of the error here.

Idea or request for content:

Make it clearer that the target values ("observed") should be 0-based, like in regular Python. Similar change could also be made for OrderedProbit.

[ricardoV94]

Sounds good @Mithrillion , wanna push a fix?

[androemeda]

@ricardoV94 please assign it to me , i would like to fix it

[charanbhatia]

@Mithrillion can you assign this issue to me?

[kdotmanoj]

@Mithrillion @ricardoV94 Could you please review PR #7457 related to the issue?