Documentation for Distribution.support() lacking

[ab-10]

Looking at the documentation I'm unable to understand for what distributions is support() defined.

For example the following works:

var b = Bernoulli({p: 0.5})
print("Possible values sampled from Bernoulli: " + b.support())

>>> Possible values sampled from Bernoulli: true,false

whereas the same on Gaussian distribution fails:

var g = Gaussian({mu: 5, sigma: 4})
print("Possible values sampled from Gaussian: " + g.support())
>>> g.support is not a function

[null-a]

My recollection is that we use support in two main ways:

1. Distributions with finite support always have a support method that returns an array containing all elements over which the distribution is defined.
2. Distributions that have support over a bounded subset of the real line use support to indicate lower and/or upper bounds.

Other distributions (e.g. those with countable support, support over entire real line, multivariate distributions) likely don't implement the method.

Yes, this isn't documented, in-part because it's not clear this is part of the public API I think. I'll close this as a duplicate of #894, but happy to try to answer further questions if you have them.

[ab-10]

Thanks for the quick response, I'll redirect further conversation to the #894.

…

On Tue, Dec 8, 2020, 11:16 AM null-a ***@***.***> wrote: My recollection is that we use support in two main ways: 1. Distributions with finite support always have a support method that returns an array containing all elements over which the distribution is defined. 2. Distributions that have support over a bounded subset of the real line use support to indicate lower and/or upper bounds. 3. Other distributions (e.g. those with countable support, support over entire real line, multivariate distributions) likely don't implement the method. Yes, this isn't documented, in-part because it's not clear this is part of the public API I think. I'll close this as a duplicate of #894 <#894>, but happy to try to answer further questions if you have them. — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub <#937 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AC54L5W2TGJ3ASP2VZP7CXLSTYDJ5ANCNFSM4URYN67Q> .