Systematic testing

[DominiqueMakowski]

This night I was thinking about a more systematic and convenient way of testing our functions, and how it was tedious to come up with particular models, fitting them every time in tests and so on.

Then I had the idea of the circus package, basically a repo that would be used only for storing all varieties and particular cases of models. We could just download this package in testing, and run the tests on all the models to see if the functions behave as expected. This could also help with the time-related issues for testing Bayesian models, as the fitting would already be done. We could quickly cover all the models, fitted once and for all.

Do you think this would work?

[strengejacke]

Having pre-compiled models would be great, indeed. But why would you put this into a package rather than in a RData file?

[DominiqueMakowski]

Stan models are quite heavy, so it seemed easier having them stored aside (clearly findable and reproducible), rather than crammed into the actual packages...

[strengejacke]

Yeah, make sense. We should think about it, though I'm not sure at the moment how to implement this properly.

[strengejacke]

I think this can be closed, right?