Eric Ward

Are there any issues to prioritize as critical before sending to CRAN? Happy to tackle some -- looking at these as being maybe higher on the list? Dispersion formula https://github.com/pbs-assess/sdmTMB/issues/70...

Sort of related, it would also be helpful to add an error message to `predict()` when `re_form_iid` is not specified -- but there are new levels in `newdata`. Without the...

Thanks, I implemented the sparse matrix approach -- which makes more sense. brms() takes a slightly different approach, passing in eigenvectors as data -- but that's just because the sparse...

Awesome - thanks for doing that. I'll simulation test it a bit more. The sum to zero constraint is mostly an issue for the ICAR models (alpha = 1). The...

I should have realized this -- but I think the issue before was the magnitude of spatial variance in the simulated data was a bit too small. Tried some other...

Just a note on where this is at: CAR model seems to be working as expected, and I've added optional beta priors on alphas. Also added better documentation for the...

I'm just seeing this thread, and confirmed the example still isn't working. Looks like the family argument is not being parsed correctly -- and crashes here https://github.com/pbs-assess/sdmTMB/blob/aad58e0a11f93abd987aca68a52d820b87895f56/R/fit.R#L554 Not perfect, but...

Confirming that using the individual random effect is probably not a good way to go, and not identifiable. Checking with assessors on how their model differs ``` data("pcod_2011") mesh

Hey @jindivero nice work here. I thought this looks pretty complete. A couple things on my to do list: - given we're using the `bei` data in a couple places,...

Never mind my point about the knot locations. I forgot that knots can only be specified to gam() -- not s(). It's possible to create dummy data beyond the range...