Handle rank-deficient LinearModel. Also changes in docs #207
Conversation
src/GLM.jl
Outdated
| @@ -76,8 +76,18 @@ module GLM | |||
| const FP = AbstractFloat | |||
| const FPVector{T<:FP} = AbstractArray{T,1} | |||
|
|
|||
| """ | |||
There was a problem hiding this comment.
Wrong indentation. Same below.
There was a problem hiding this comment.
What is the correct indentation in this case? I thought that the first line within the multi-line string should be indented 4 spaces (b/c Markdown) and the line beginning with "Abstract" should not be indented. If I indent the triple double-quote characters things will look weird.
Perhaps the best solution is to move those abstract type declarations to linpred.jl and lm.jl.
There was a problem hiding this comment.
No, you can use whatever indentation you want (i.e. the same as surrounding lines), what matters is the alignment with the """.
src/glmfit.jl
Outdated
| @@ -54,6 +54,8 @@ When `L` is the canonical link for `D` the derivative of the inverse link is a m | |||
| of the variance function for `D`. If they are the same a numerator and denominator term in | |||
| the expression for the working weights will cancel. | |||
| """ | |||
| function cancancel end | |||
There was a problem hiding this comment.
Any reason not to attach the docstring to the method below? If that's because it doesn't appear in the manual, I think you to list argument types there, i.e. cancancel(::GlmResp).
test/runtests.jl
Outdated
| @testset "rankdeficient" begin | ||
| # an example of rank deficiency caused by a missing cell in a table | ||
| dfrm = DataFrame([categorical(repeat(string.('A':'D'), inner = 6)), | ||
| categorical(repeat(string.('a':'c'), inner = 2, outer = 4))], |
test/runtests.jl
Outdated
| ymissingcell = y[inds] | ||
| @test_throws LinAlg.PosDefException m2 = fit(LinearModel, Xmissingcell, ymissingcell) | ||
| m2p = fit(LinearModel, Xmissingcell, ymissingcell, true) | ||
| @test isapprox(deviance(m2p), 0.2859221258731563) |
There was a problem hiding this comment.
Wouldn't it make sense to test the values of coefficients?
|
|
||
| makedocs() | ||
| makedocs( | ||
| format = :html, |
There was a problem hiding this comment.
I think you can remove docs/mkdocs.yml with that option. And of course the old docs/ folder.
For reasons that I am having difficulty remembering, I incorporated documentation changes in these changes to handle rank-deficient linear models as discussed in #200.
I realize this makes it harder to review these changes for rank-deficiency. I can start a new branch and cherry-pick the rank-deficiency-related commits if that would be preferred.