AbstractReTerm

[palday]

Closes #293. Not sure if this is a good or bad idea. Thoughts, @kleinschmidt ?

[codecov]

Codecov Report

Merging #395 into master will not change coverage.
The diff coverage is n/a.

@@           Coverage Diff           @@
##           master     #395   +/-   ##
=======================================
  Coverage   95.32%   95.32%           
=======================================
  Files          23       23           
  Lines        1605     1605           
=======================================
  Hits         1530     1530           
  Misses         75       75           

Impacted Files	Coverage Δ
src/randomeffectsterm.jl	94.11% <ø> (ø)

---

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 5d3988d...ef67e39. Read the comment docs.

[dmbates]

Should FullDummy also be an AbstractReTerm?

[palday]

@dmbates Good question. I'm hoping @kleinschmidt has a strong feeling on that, but I'll give it a think. The only downside I see is that StatsModels also has fulldummy for any categorical predictor and fulldummy is defined at the level of a single categorical experimental variable, while zerocorr is approximately at the level of a grouping variable.

[kleinschmidt]

If this is just for dispatch then it seems harmless to me. There's no struct to represent full dummy coding (it just creates a new CategoricalTerm using FullDummyCoding) so that's not an issue. RandomEffectsTerm doesn't seem to be used elsewhere in the code (e.g. to restrict what can be stored as the reterms in the model struct) so no objections there either.

[dmbates]

Other than the URL change, it looks good.

There is a script in the docs/ directory to create the footnotes automatically.