Douglas Bates

I think that will work. We could try to make it easier to construct a `Generalized Linear Mixed Model` from the `feterm` and `reterms` but I am a bit reluctant...

@MaximilianNuber Is it correct that there is only one random-effects term in the LMMs and GLMMs that you wish to fit and that it is of the form `(1|samp)`? If...

@andreasnoack Have you tried the PRIMA versions of BOBYQA and COBYLA? There is an extension to MixedModels.jl that is available if you add PRIMA first in your session. It provides...

I agree we need to look into this in greater detail. The issue @ajinkya-k mentioned is that editors trying to reproduce results from a paper we submitted got substantially different...

Another issue for us is the use of accelerated BLAS, such as AppleAccelerate or MKL. When you have two or more grouping factors for the random effects and large numbers...

I tried the PRIMA versions of BOBYQA and COBYLA under Julia 1.10.9 with essentially the same results - convergence to an objective of 237648.6016 after 164 evaluations for BOBYQA and...

I tried to recreate the premature termination with Julia-1.11.5 on an M-series MacOS laptop and was unable to ```julia julia> using DataFrames, CSV, MixedModels, CairoMakie julia> df = CSV.read("issues/833/data.csv", DataFrame)...

 This is the plot of the objective for this model.

I have created a `db/issue833` branch with the data and the script I used in an `issues/833` directory.

With MixedModels v5.0.0 and later the default optimizer, which is currently :LN_BOBYQA (without lower bounds) for 1-dimentional problems, seems to work fine on Mac and Linux. @andreasnoack Do you think...