Phillip Alday
Phillip Alday
There are two issues here: 1. recovery from failed profiling 2. failure of profiling. The first is easier to fix and I'll something addressing it in the next few days.
> I don't know why "threshold=3" gives the error, but "threshold=4" appears to work. I think this is an artifact of execution order. If you go to `threshold=4` in a...
Okay, very weird, I get the error on `threshold=3` but not `threshold=4`. But `threshold=3` should be part of `threshold=4`. 🤔
I think I understand the underlying problem with `threshold=3`: the cubic spline that we fit to the profiled points for interpolation just doesn't work without the extra data that `threshold=4`...
It might be technically breaking to not store the sigma parameter, but we can definitely improve display and handling of it.
If it can wait a few days, I can do it. If you have big plans for it this week, then you should do it. :smile:
> Do we have some volunteers to help make the release on the CRAN side? maybe @Non-Contradiction ? I will also have a bit more time on Friday (15 November).
@andreasnoack Can you show us `versioninfo()`? @dmbates has actually done a fair number of more involved experiments with using the gradient and it never really paid off. That said, we've...
> The reason I had the ForwardDiff compatible deviance code handy is actually that I'm using it for the Satterthwaite approximation for the degrees of freedom. I'd be happy to...
@andreasnoack it's stale relative to `main`, but I think this was the most recent effort: https://github.com/JuliaStats/MixedModels.jl/tree/forwarddiff