Allow for offset in GLMMs and simplify wts

[dmbates]

• closes Offset for Poisson models #453
• working on an example based on one in lme4

[codecov]

Codecov Report

Merging #482 (d682d28) into master (135e87d) will increase coverage by 0.65%.
The diff coverage is 100.00%.

❗ Current head d682d28 differs from pull request most recent head 2474373. Consider uploading reports for the commit 2474373 to get more accurate results

@@            Coverage Diff             @@
##           master     #482      +/-   ##
==========================================
+ Coverage   93.36%   94.01%   +0.65%     
==========================================
  Files          27       27              
  Lines        1929     2140     +211     
==========================================
+ Hits         1801     2012     +211     
  Misses        128      128              

Impacted Files	Coverage Δ
src/generalizedlinearmixedmodel.jl	89.16% <100.00%> (+0.99%)	⬆️
src/linearmixedmodel.jl	95.44% <100.00%> (+0.43%)	⬆️
src/likelihoodratiotest.jl	97.64% <0.00%> (-1.04%)	⬇️
src/blocks.jl	100.00% <0.00%> (ø)
src/varcorr.jl	100.00% <0.00%> (ø)
src/datasets.jl	100.00% <0.00%> (ø)
src/utilities.jl	100.00% <0.00%> (ø)
src/optsummary.jl	100.00% <0.00%> (ø)
src/gausshermite.jl	100.00% <0.00%> (ø)
src/linalg/pivot.jl	100.00% <0.00%> (ø)
... and 17 more

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 135e87d...2474373. Read the comment docs.

[dmbates]

@palday Should we go ahead and merge this, perhaps after turning off the Future ci. I would say that this is a bug fix. (I suppose now you want me to put it in News.md, don't you?)

[palday]

@dmbates you predicted my request. 😄 Can you put it under 3.4.1 since it's technically a bug fix?

As a sidebar: should we support offsets in LMM (in a separate PR)?

[dmbates]

This branch is currently based on the master branch because I was intending it to check off one of the milestones for v4.0.0, I know it is a bug fix and could be incorporated as v3.4.1 but I am not sure my git skills are up to backing out the other changes in the master branch.

[palday]

Put the NEWS entry in the right spot, we'll merge into master and I'll git-magic the backport aftwards. 😄

[dmbates]

As a sidebar: should we support offsets in LMM (in a separate PR)?

The uses of an offset for a model with identity link have always seemed rather contrived to me. You might as well just shift the response by the offset and proceed normally.

[palday]

The uses of an offset for a model with identity link have always seemed rather contrived to me. You might as well just shift the response by the offset and proceed normally.

My thoughts exactly. 😄 Should we error if an offset is specified for an LMM then?

[dmbates]

The uses of an offset for a model with identity link have always seemed rather contrived to me. You might as well just shift the response by the offset and proceed normally.

My thoughts exactly. smile Should we error if an offset is specified for an LMM then?

Yes, please do.

[palday]

The constructor clearly will error, but for the fit(MixedModel, formula, data; offset=data.o) I don't know what will happen.

I'll add it later today and then get this merged in. 😄