Test of bivariate uniformity

[StaffanBetner]

An overall test of independence of errors could be a test of bivariate uniformity of the rank transformed predicted values versus standardized residuals. Maybe an idea to implement this in DHARMa? (And maybe do some simulations to see if the approach is good)

[florianhartig]

Hi Staffan, thanks for the comment. I think what you propose is what I describe in issue https://github.com/florianhartig/DHARMa/issues/72 (sorry, it's super short, was just a reminder to me. My idea was to fit a homogenous Poisson model to the rank-transformed standardised residuals).

Is that what you had in mind? Or is there another test for bivariate uniformity that could work better?

[florianhartig]

Hi @StaffanBetner - any more comments on that? Otherwise I will close this as a duplicate of #72.

[StaffanBetner]

@florianhartig There are a couple of available tests for bivariate uniformity. A couple of them are implemented in the R package SHT. Although I do not now how a test of bivariate uniformity would be equal to a test of homogenous poisson process, or which approach that is best suited.

[florianhartig]

OK, thanks, I'll have a look at that!