Math notation for models?

[eerolinna]

Would it be good to include math notation for models?

For blr model this would mean having something like

\beta \sim \mathcal{N}(0,10) \\
\sigma \sim \mathcal{N}(0, 10) \\
y \sim \mathcal{N}(X\beta, \sigma)

somewhere in the model info file.

I will say straight away that I'm totally fine with not adding this if we feel like this is unnecessary. In that case feel free to close this.

I guess the biggest use for this would be that if we have a website that has a page for each model/dataset/posterior we could render the math notation there.

Another use could be that currently the statistical relationships between variables are defined implicitly by implementation*, having math notation would give an explicit definition. However I guess a caveat is that the math notation is not always precise enough, which would mean that we would still sometimes only have an implicit definition by implementation.

Again, if this doesn't feel like something that would be worth the effort feel free to close this.

[*] What I mean by this is that we can have many implementations of the model (say stan and pymc) and the way we can say that an implementation is correct is if

• for the first implementation, it implements the model we "meant" it to implement. However what a model is supposed to be is only in our heads
• for other implementations, if it is the same as the first implementation (or a different wording: if all implementations implement the same model)

[MansMeg]

I have had this in mind myself multiple times. I think this is a really good idea. The only problem is how to include more strange models where we don't have nice math notations. See some of the models Oliver has added.

[eerolinna]

I guess one option would be to just omit the math notation for those models, in other words make the math notation recommended but not mandatory.