CompatHelper: bump compat for "StatsModels" to "0.7" (#62)

* CompatHelper: bump compat for "StatsModels" to "0.7" * fixes, compat re-arranging * set progress=false in tests * YASG --------- Co-authored-by: Phillip Alday <[email protected]>
palday · Apr 14, 2023 · 7b3a29f · 7b3a29f · palday · Apr 14, 2023
1 parent cea7a0c
commit 7b3a29f
Show file tree

Hide file tree

Showing 8 changed files with 59 additions and 43 deletions.
diff --git a/Project.toml b/Project.toml
@@ -1,7 +1,7 @@
 name = "JellyMe4"
 uuid = "19ac8677-9a15-4623-9afd-84acc6165ce7"
 authors = ["Phillip Alday <[email protected]>"]
-version = "0.2.7"
+version = "0.2.8"
 
 [deps]
 CategoricalArrays = "324d7699-5711-5eae-9e2f-1d82baa6b597"
@@ -17,16 +17,16 @@ StatsModels = "3eaba693-59b7-5ba5-a881-562e759f1c8d"
 Tables = "bd369af6-aec1-5ad0-b16a-f7cc5008161c"
 
 [compat]
-CategoricalArrays = "0.7, 0.8, 0.9, 0.10"
-DataFrames = "0.20, 0.21, 0.22, 1.0"
-Distributions = "0.20, 0.22, 0.23, 0.24, 0.25"
+CategoricalArrays = "0.10"
+DataFrames = "1"
+Distributions = "0.25"
 GLM = "1.3"
-MixedModels = "3.0, 4.0"
+MixedModels = "4"
 RCall = "0.13"
 RegressionFormulae = "0.1"
 StatsBase = "0.31, 0.32, 0.33"
-StatsModels = "0.6.8"
-Tables = "0.2, 1.0"
+StatsModels = "0.6, 0.7"
+Tables = "1"
 julia = "1.6"
 
 [extras]

diff --git a/deps/build.jl b/deps/build.jl
@@ -1 +1 @@
-# should we try to install lme4 or even afex automatically at build time?
+# should we try to install lme4 or even afex automatically at build time?
diff --git a/src/glmerMod.jl b/src/glmerMod.jl
@@ -119,7 +119,7 @@ function rcopy(::Type{GeneralizedLinearMixedModel}, s::Ptr{S4Sxp})
 
     if length(θ) != length(m.θ)
         @error """You're probably using || in R with a categorical variable,
-                  whose translation is currently unsupported with MixedModels 3.0."""
+                  whose translation is currently unsupported with recent MixedModels releases"""
         throw(ArgumentError("Parameter vectors in R and Julia are different sizes."))
     end
 

diff --git a/src/lmerMod.jl b/src/lmerMod.jl
@@ -57,7 +57,7 @@ function rcopy(::Type{LinearMixedModel}, s::Ptr{S4Sxp})
 
     if length(θ) != length(m.θ)
         @error """You're probably using || in R with a categorical variable,
-                  whose translation is currently unsupported with MixedModels 3.0."""
+                  whose translation is currently unsupported with recent MixedModels releases"""
         throw(ArgumentError("Parameter vectors in R and Julia are different sizes."))
     end
 

diff --git a/src/utilities.jl b/src/utilities.jl
@@ -41,16 +41,30 @@ function get_r_contrasts(rdf)
     data = rcopy(rdf)
     # get categorical columns
     cnames = [c for c in propertynames(data) if typeof(data[!, c]) <: CategoricalArray]
-    return Dict(c => HypothesisCoding(pinv(rcopy(R"contrasts($(rdf[c]))"));
-                                      labels=rcopyarray(R"colnames(contrasts($(rdf[c])))"))
-                for c in cnames)
+    pairs = []
+    for c in cnames
+        @debug "" c
+        levels = rcopyarray(R"rownames(contrasts($(rdf[c])))")
+        @debug "" levels
+        if levels == [nothing]
+            # for some contrasts, the rownames aren't defined
+            # we don't prefer this method because it's possible
+            # that it's not quite right when people are using really
+            # fancy contrasts and it can fail
+            levels = rcopyarray(R"levels($(rdf[c]))")
+        end
+        labels = rcopyarray(R"colnames(contrasts($(rdf[c])))")
+        @debug "" labels
+        push!(pairs,
+              c => HypothesisCoding(pinv(rcopy(R"contrasts($(rdf[c]))")); levels, labels))
+    end
+    return Dict(pairs)
 end
 
 """
 set the contrasts on an R dataframe
 this of course has a side effect: it changes the R dataframe
 """
-
 function set_r_contrasts!(rdfname, formula)
     fixefform = first(formula.rhs)
 

diff --git a/test/glmerMod.jl b/test/glmerMod.jl
@@ -132,7 +132,7 @@ logistic(x) = 1 / (1 + exp(-x))
                 # we max out at 1 scalar RE for the nAGQ tests
                 jlmm = GLMM(@formula(r2 ~ 1 + anger + gender + btype + situ + (1 | subj)),
                             dat, Bernoulli())
-                fit!(jlmm; fast=false, nAGQ=9)
+                fit!(jlmm; fast=false, nAGQ=9, progress=false)
                 jm = (jlmm, dat)
                 @rput jm
                 # @test_warn Regex(".*categorical.*") @rput jm;
@@ -147,7 +147,7 @@ logistic(x) = 1 / (1 + exp(-x))
                 # we max out at 1 scalar RE for the nAGQ tests
                 jlmm = GLMM(@formula(r2 ~ 1 + anger + gender + btype + situ + (1 | subj)),
                             dat, Bernoulli())
-                fit!(jlmm)
+                fit!(jlmm; progress=false)
                 jm = Tuple([jlmm, dat])
                 @rput jm
 
@@ -159,7 +159,7 @@ logistic(x) = 1 / (1 + exp(-x))
             @testset "columntable" begin
                 jlmm = GLMM(@formula(r2 ~ 1 + anger + gender + btype + situ + (1 | subj)),
                             dat, Bernoulli())
-                fit!(jlmm)
+                fit!(jlmm; progress=false)
                 jm = Tuple([jlmm, columntable(dat)])
                 @rput jm
             end
@@ -170,7 +170,7 @@ logistic(x) = 1 / (1 + exp(-x))
             dat = dataset(:cbpp)
             dat.rate = dat.incid ./ dat.hsz
             jlmm = fit(MixedModel, @formula(rate ~ 1 + period + (1 | herd)),
-                       dat, Binomial(); wts=float(dat.hsz), fast=true)
+                       dat, Binomial(); wts=float(dat.hsz), fast=true, progress=false)
 
             jm = (jlmm, dat)
             @rput jm
@@ -192,7 +192,7 @@ logistic(x) = 1 / (1 + exp(-x))
                         dat, Poisson())
 
             jm = (jlmm, dat)
-            fit!(jlmm; fast=true) # problems with this one in fast=false
+            fit!(jlmm; fast=true, progress=false) # problems with this one in fast=false
             @rput jm
             # @test_warn Regex(".*categorical.*") @rput jm;
             @test rcopy(R"""jm@devcomp$dims["nAGQ"]""") == 0
@@ -228,7 +228,7 @@ logistic(x) = 1 / (1 + exp(-x))
             jlmm = fit(MixedModel,
                        @formula(r2 ~ 1 + anger + gender + btype + situ + (1 | subj) +
                                      (1 | item)),
-                       dat, Bernoulli();
+                       dat, Bernoulli(); progress=false,
                        contrasts=Dict(:gender => EffectsCoding(),
                                       :btypes => EffectsCoding()))
             jm = (jlmm, dat)

diff --git a/test/lmerMod.jl b/test/lmerMod.jl
@@ -1,9 +1,10 @@
-using RCall, MixedModels, Test
+using RCall, JellyMe4, MixedModels, Test
 using StatsBase: zscore
 using StatsModels: SeqDiffCoding
 using Tables: columntable
 
-import JellyMe4: _set_lmer, _set_afex_installed
+using JellyMe4: _set_lmer, _set_afex_installed
+using MixedModels: dataset
 
 const LMM = LinearMixedModel
 const GLMM = GeneralizedLinearMixedModel
@@ -25,22 +26,22 @@ const GLMM = GeneralizedLinearMixedModel
 
     # this is available in MixedModels.dataset(:sleepstudy) but with different
     # capitalization than in R
-    sleepstudy = rcopy(R"sleepstudy")
+    sleepstudy = rcopy(R"lme4::sleepstudy")
 
     kb07 = dataset(:kb07)
 
     @testset "get lmerMod" begin
         ### from R ###
 
         jlmm = fit!(LMM(@formula(Reaction ~ 1 + Days + (1 + Days | Subject)), sleepstudy);
-                    REML=false)
-        rlmm = rcopy(R"m <- lme4::lmer(Reaction ~ 1 + Days + (1 + Days|Subject),sleepstudy,REML=FALSE)")
+                    REML=false, progress=false)
+        rlmm = rcopy(R"m <- lme4::lmer(Reaction ~ 1 + Days + (1 + Days|Subject),lme4::sleepstudy,REML=FALSE)")
 
         @test jlmm.θ ≈ rlmm.θ atol = 0.001
         @test objective(jlmm) ≈ objective(rlmm) atol = 0.001
         @test fixef(jlmm) ≈ fixef(rlmm) atol = 0.001
 
-        jlmm = refit!(jlmm; REML=true)
+        jlmm = refit!(jlmm; REML=true, progress=false)
         rlmm = rcopy(R"update(m, REML=TRUE)")
 
         @test jlmm.θ ≈ rlmm.θ atol = 0.001
@@ -49,14 +50,14 @@ const GLMM = GeneralizedLinearMixedModel
 
         @testset "merModLmerTest" begin
             jlmm = fit!(LMM(@formula(Reaction ~ 1 + Days + (1 + Days | Subject)),
-                            sleepstudy); REML=false)
+                            sleepstudy); REML=false, progress=false)
             rlmm = rcopy(R"m <- lmerTest::lmer(Reaction ~ 1 + Days + (1 + Days|Subject),sleepstudy,REML=FALSE)")
 
             @test jlmm.θ ≈ rlmm.θ atol = 0.001
             @test objective(jlmm) ≈ objective(rlmm) atol = 0.001
             @test fixef(jlmm) ≈ fixef(rlmm) atol = 0.001
 
-            jlmm = refit!(jlmm; REML=true)
+            jlmm = refit!(jlmm; REML=true, progress=false)
             rlmm = rcopy(R"update(m, REML=TRUE)")
 
             @test jlmm.θ ≈ rlmm.θ atol = 0.001
@@ -70,7 +71,7 @@ const GLMM = GeneralizedLinearMixedModel
         # the scalar-vector distinction for θ is missing in R
         @testset "scalar RE" begin
             jlmm = fit!(LMM(@formula(Reaction ~ 1 + Days + (1 | Subject)), sleepstudy);
-                        REML=false)
+                        REML=false, progress=false)
             rlmm = rcopy(R"m <- lme4::lmer(Reaction ~ 1 + Days + (1|Subject),sleepstudy,REML=FALSE)")
 
             @test jlmm.θ ≈ rlmm.θ atol = 0.001
@@ -91,7 +92,7 @@ const GLMM = GeneralizedLinearMixedModel
 
         @testset "contrasts" begin
             reval("""
-            data(cake)
+            cake <- lme4::cake
             cake\$rr <- with(cake, replicate:recipe)
             """)
             rlmm = rcopy(R"fm1 <- lme4::lmer(angle ~ recipe * temperature + (1|rr), cake, REML= FALSE)")
@@ -103,7 +104,7 @@ const GLMM = GeneralizedLinearMixedModel
 
         @testset "caret" begin
             reval("""
-            data(cake)
+            cake <- lme4::cake
             cake\$rr <- with(cake, replicate:recipe)
             """)
             rlmm = rcopy(R"fm1 <- lme4::lmer(angle ~ (recipe + temperature)^2 + (1|rr), cake, REML= FALSE)")
@@ -130,7 +131,7 @@ const GLMM = GeneralizedLinearMixedModel
 
             rlmm = rcopy(R"m <- lme4::lmer(Reaction ~ 1 + Days + (1 + Days||Subject),sleepstudy,REML=FALSE)")
             jlmm = fit!(LMM(@formula(Reaction ~ 1 + Days + zerocorr(1 + Days | Subject)),
-                            sleepstudy); REML=false)
+                            sleepstudy); REML=false, progress=false)
             # as a cheat for comparing the covariance matrices, we use PCA
             @test only(rlmm.rePCA) ≈ only(jlmm.rePCA) atol = 0.05
         end
@@ -151,7 +152,7 @@ const GLMM = GeneralizedLinearMixedModel
         @test rcopy(R"fitted(jm)") ≈ fitted(jlmm)
         @test rcopy(R"REMLcrit(jm)") ≈ objective(jlmm)
 
-        refit!(jlmm; REML=false)
+        refit!(jlmm; REML=false, progress=false)
         jm = (jlmm, sleepstudy)
         @rput jm
         @test rcopy(R"fitted(jm)") ≈ fitted(jlmm)
@@ -168,14 +169,14 @@ const GLMM = GeneralizedLinearMixedModel
             R"m <- lme4::lmer(log10(Reaction) ~ 1 + log(Days2) + (1 + log(Days2)|Subject),sleepstudy,REML=FALSE)"
             jlmm = fit!(LMM(@formula(log10(Reaction) ~ 1 + log(Days2) +
                                                        (1 + log(Days2) | Subject)),
-                            sleepstudy); REML=false)
+                            sleepstudy); REML=false, progress=false)
             jm = (jlmm, sleepstudy)
             @rput jm
             @test rcopy(R"fitted(jm)") ≈ fitted(jlmm)
             @test rcopy(R"deviance(jm)") ≈ objective(jlmm)
 
             jlmm = fit!(LMM(@formula(Reaction ~ 1 + round(Days) + (1 | Subject)),
-                            sleepstudy); REML=false)
+                            sleepstudy); REML=false, progress=false)
             jm = (jlmm, sleepstudy)
 
             @test_throws ArgumentError (@rput jm)
@@ -184,7 +185,7 @@ const GLMM = GeneralizedLinearMixedModel
         @testset "sorting by n BLUPs vs. n groups" begin
             jlmm = fit(LMM,
                        @formula(rt_trunc ~ 1 + (1 | subj) + (1 + load + prec + spkr | item)),
-                       kb07)
+                       kb07; progress=false)
             jm = (jlmm, kb07)
             @rput jm
             @test rcopy(R"fitted(jm)") ≈ fitted(jlmm)
@@ -202,7 +203,8 @@ const GLMM = GeneralizedLinearMixedModel
             cake
             """))
             jlmm = fit(MixedModel, @formula(angle ~ recipe * temperature + (1 | rr)),
-                       cake; REML=false, contrasts=Dict(:temperature => SeqDiffCoding()))
+                       cake; REML=false, progress=false,
+                       contrasts=Dict(:temperature => SeqDiffCoding()))
             jm = (jlmm, cake)
             @rput jm
             @test fixef(jlmm) ≈ rcopy(R"fixef(jm)")
@@ -212,7 +214,7 @@ const GLMM = GeneralizedLinearMixedModel
             machines = rcopy(R"as.data.frame(nlme::Machines)")
             jlmm = fit(MixedModel,
                        @formula(score ~ 1 + Machine + (1 + fulldummy(Machine) | Worker)),
-                       machines)
+                       machines; progress=false)
             rlmm = (jlmm, machines)
             @rput rlmm
             rlmmrepca = rcopy(R"summary(rePCA(rlmm))$Worker$importance[3,]")
@@ -227,7 +229,7 @@ const GLMM = GeneralizedLinearMixedModel
                 _set_afex_installed(false)
 
                 jlmm = fit!(LMM(@formula(Reaction ~ 1 + Days + zerocorr(1 + Days | Subject)),
-                                sleepstudy); REML=false)
+                                sleepstudy); REML=false, progress=false)
                 rlmm = (jlmm, sleepstudy)
                 @rput rlmm
                 @test rcopy(R"""!is(rlmm,"merModLmerTest")""")
@@ -241,7 +243,7 @@ const GLMM = GeneralizedLinearMixedModel
 
                 jlmm = fit(MixedModel,
                            @formula(score ~ 1 + Machine + zerocorr(0 + Machine | Worker)),
-                           machines)
+                           machines; progress=false)
 
                 @testset "afex pre-enabled" begin
                     _set_lmer("afex::lmer_alt")
@@ -279,7 +281,7 @@ const GLMM = GeneralizedLinearMixedModel
         slp[!, :obs] .= 1:nrow(slp)
         fm1 = fit(MixedModel,
                   @formula(reaction ~ 1 + days + (1 | obs)),
-                  slp;
+                  slp; progress=false,
                   contrasts=Dict(:obs => Grouping()))
         rfm1 = (fm1, slp)
         warning = r"""Warning: number of levels of each grouping factor must be < number of observations \(problems: obs\)

diff --git a/test/merMod.jl b/test/merMod.jl
@@ -6,9 +6,9 @@ const GLMM = GeneralizedLinearMixedModel
 @testset "merMod" begin
     # this is available in MixedModels.dataset(:sleepstudy) but with different
     # capitalization than in R
-    sleepstudy = rcopy(R"sleepstudy")
+    sleepstudy = rcopy(R"lme4::sleepstudy")
     jlmm = fit!(LMM(@formula(Reaction ~ 1 + round(Days) + (1 | Subject)), sleepstudy);
-                REML=false)
+                REML=false, progress=false)
     @testset "bare model" begin
         @test_throws ArgumentError (@rput jlmm)
     end