Phillip Alday

This looks like two questions. 1. Regarding your first question on nesting: the example you have works for me without error. Testing of the nesting is somewhat sensitive to floating...

I think we're talking past each other: - Your F-test MWE works for me and does not error. - Your example of `dropcollinear` does exactly what is documented, i.e. sets...

Can you post your `versioninfo()` and `using Pkg; Pkg.status()` so that I can see all the relevant architecture and version details? Note that nesting and rank deficiency is one of...

The problem is definitely numerical instability. I have now tested this example on: - Julia 1.7 with OpenBLAS on Linux x64 (11th Gen Intel(R) Core(TM) i5-1135G7 @ 2.40GHz) - Julia...

@nalimilan completely agree on the QR decomposition. I wanted to mock up how to do it for this example, but it was actually not nearly as polished as the Cholesky...

The implementation in MixedModelsExtras will actually already handle (some) models fit with GLM.jl: https://github.com/palday/MixedModelsExtras.jl/blob/d92910c011ca4c920a4e5e7dfc84b99d87f12754/test/vif.jl#L45-L61 ```julia @testset "GVIF and RegrssionModel" begin duncan = rdataset("car", "Duncan") lm1 = lm(@formula(Prestige ~ 1 +...

@bkamins The code is written; there is just a little bit of discussion about what the API should look like and where various parts should live: https://github.com/JuliaStats/MixedModels.jl/issues/661#issuecomment-1368523710 Currently we have...

if @kleinschmidt is happy for StatsModels to own `termnames`, then I'll open a PR there (against both 0.6 and 0.7 series), get that merged, then I'll move `vif` over to...

@HiramTheHero in the meantime, you should "just" be able to do ```julia using MixedModelsExtras vif(my_glm_model) ``` and things will probably work.

We are not responsible for how users use or misuse a particular statistical quantity -- otherwise, I certainly wouldn't want to be responsible for any code that outputs a p-value!...