Talk:Conditional probability table

Marginal or conditional?

I wonder, why conditional probabilities are called "marginal" in this article (twice)?

I also wonder, is it notable? Whenever something is defined (and discrete), one may tabulate it, making a table; so what? And if at all this deserves mentioning, why a separate page? Why not a line or two in Conditional probability distribution? There is not enough content for a separate article.

Boris Tsirelson (talk) 05:37, 2 June 2014 (UTC)[reply]

CPTs - page justification

I do think this article merits a separate page, though more content is needed. When using Bayesian networks to calculate joint probability distributions (studying this now in my AI class), conditional probability tables (CPTs) are necessary and are commonly referenced in Bayesian network literature. People may come across the term CPT in ML/AI/statistics literature, and Wikipedia should have a landing page for it in my opinion. Also, CPTs have certain interesting properties that are not yet well covered in this page. For example, this page should probably cover how to marginalize out a random variable from a CPT.

Conditional and marginal probabilities are related but are not the same - I don't think the article is calling conditional probability "marginal", but I think it's defining the relationship between the two. The definition from Murphy is admittedly dense - the Murphy book is for an advanced graduate audience, so it should be paraphrased in more readable terms. However, I would prefer someone more qualified to do the paraphrasing.

I think this page needs some fleshing out and some diagram(s).

Puhfyn (talk) 07:03, 2 June 2014 (UTC)[reply]

Then the article should say much more than just "a conditional probability table is a table of conditional probabilities". Boris Tsirelson (talk) 09:22, 2 June 2014 (UTC)[reply]