update MG tests

Artifacts.toml

@@ -15,12 +15,12 @@ lazy = true
     url = "https://github.com/StructuralEquationModels/Data/raw/0462a36018c71b8559dc76819d56a538857593c8/datasets/holzinger_swineford/compressed/data_fiml.csv.tar.gz"
 
 [holzinger_swineford_solution]
-git-tree-sha1 = "736fd9ef52635a38a420536bd5d32848329785b1"
+git-tree-sha1 = "e2ba8a13e85ea1eeaa1353bc54c9e6bf915dea3b"
 lazy = true
 
     [[holzinger_swineford_solution.download]]
-    sha256 = "7bd5132038f27ccb6f8fb318850613eae8d7f98d0e07738e64c0715a118745ed"
-    url = "https://github.com/StructuralEquationModels/Data/raw/c0e14988ebf0ccd370db6150f25126812f51329b/datasets/holzinger_swineford/compressed/holzinger_swineford_solution.tar.gz"
+    sha256 = "6c77146c68817302df0b28469d98ea650a158f375102bafa94dbe5945033eee2"
+    url = "https://github.com/StructuralEquationModels/Data/raw/a34d06ea80738f8d0f640ef421e310b03f88be23/datasets/holzinger_swineford/compressed/holzinger_swineford_solution.tar.gz"
 
 [political_democracy]
 git-tree-sha1 = "a02b07e3523570d8a27c9fe10dcac3a1e4705c33"

test/examples/examples.jl

@@ -2,6 +2,4 @@ using Test, SafeTestsets
 
 @safetestset "Political Democracy" begin include("political_democracy/political_democracy.jl") end
 @safetestset "Recover Parameters" begin include("recover_parameters/recover_parameters_twofact.jl") end
-
-#@safetestset "Multigroup" begin include("multigroup.jl") end
-#@safetestset "Multigroup Parser" begin include("multigroup_parser.jl") end
+@safetestset "Multigroup" begin include("multigroup/multigroup.jl") end

test/examples/multigroup/build_models.jl

@@ -0,0 +1,239 @@
+####################################################################
+# ML estimation
+####################################################################
+
+model_g1 = Sem(
+    specification = specification_g1,
+    data = dat_g1,
+    imply = RAMSymbolic
+)
+
+model_g2 = Sem(
+    specification = specification_g2,
+    data = dat_g2,
+    imply = RAM
+)
+
+model_ml_multigroup = SemEnsemble(model_g1, model_g2; diff = semdiff)
+
+############################################################################
+### test gradients
+############################################################################
+
+@testset "ml_gradients_multigroup" begin
+    @test test_gradient(model_ml_multigroup, start_test; atol = 1e-9)
+end
+
+# fit
+@testset "ml_solution_multigroup" begin
+    solution = sem_fit(model_ml_multigroup)
+    update_estimate!(partable, solution)
+    @test compare_estimates(
+        partable, 
+        solution_lav[:parameter_estimates_ml]; atol = 1e-4,
+        lav_groups = Dict(:Pasteur => 1, :Grant_White => 2))
+end
+
+@testset "fitmeasures/se_ml" begin
+    solution_ml = sem_fit(model_ml_multigroup)
+    @test all(test_fitmeasures(
+        fit_measures(solution_ml), 
+        solution_lav[:fitmeasures_ml]; rtol = 1e-2, atol = 1e-7))
+
+    update_partable!(
+        partable, identifier(model_ml_multigroup), se_hessian(solution_ml), :se)
+    @test compare_estimates(
+        partable, 
+        solution_lav[:parameter_estimates_ml]; atol = 1e-3, 
+        col = :se, lav_col = :se,
+        lav_groups = Dict(:Pasteur => 1, :Grant_White => 2))
+end
+
+####################################################################
+# ML estimation - user defined loss function
+####################################################################
+
+struct UserSemML <: SemLossFunction
+    objective
+    gradient
+    hessian
+end
+
+############################################################################
+### constructor
+############################################################################
+
+UserSemML(;n_par, kwargs...) = UserSemML([1.0], zeros(n_par), zeros(n_par, n_par)) 
+
+############################################################################
+### functors
+############################################################################
+
+import LinearAlgebra: Symmetric, cholesky, isposdef, logdet, tr
+import LinearAlgebra
+
+function (semml::UserSemML)(par, F, G, H, model)
+    if G error("analytic gradient of ML is not implemented (yet)") end
+    if H error("analytic hessian of ML is not implemented (yet)") end
+
+    a = cholesky(Symmetric(model.imply.Σ); check = false)
+    if !isposdef(a)
+        semml.objective[1] = Inf
+    else
+        ld = logdet(a)
+        Σ_inv = LinearAlgebra.inv(a)
+        if !isnothing(F)
+            prod = Σ_inv*model.observed.obs_cov
+            semml.objective[1] = ld + tr(prod)
+        end
+    end
+end
+
+# models
+model_g1 = Sem(
+    specification = specification_g1,
+    data = dat_g1,
+    imply = RAMSymbolic
+)
+
+model_g2 = SemFiniteDiff(
+    specification = specification_g2,
+    data = dat_g2,
+    imply = RAMSymbolic,
+    loss = UserSemML
+)
+
+model_ml_multigroup = SemEnsemble(model_g1, model_g2; diff = semdiff)
+
+@testset "gradients_user_defined_loss" begin
+    @test test_gradient(model_ml_multigroup, start_test; atol = 1e-9)
+end
+
+# fit
+@testset "solution_user_defined_loss" begin
+    solution = sem_fit(model_ml_multigroup)
+    update_estimate!(partable, solution)
+    @test compare_estimates(
+        partable, 
+        solution_lav[:parameter_estimates_ml]; atol = 1e-4,
+        lav_groups = Dict(:Pasteur => 1, :Grant_White => 2))
+end
+
+####################################################################
+# GLS estimation
+####################################################################
+
+model_ls_g1 = Sem(
+    specification = specification_g1,
+    data = dat_g1,
+    imply = RAMSymbolic,
+    loss = SemWLS
+)
+
+model_ls_g2 = Sem(
+    specification = specification_g2,
+    data = dat_g2,
+    imply = RAMSymbolic,
+    loss = SemWLS
+)
+
+model_ls_multigroup = SemEnsemble(model_ls_g1, model_ls_g2; diff = semdiff)
+
+@testset "ls_gradients_multigroup" begin
+    @test test_gradient(model_ls_multigroup, start_test; atol = 1e-9)
+end
+
+@testset "ls_solution_multigroup" begin
+    solution = sem_fit(model_ls_multigroup)
+    update_estimate!(partable, solution)
+    @test compare_estimates(
+        partable, 
+        solution_lav[:parameter_estimates_ls]; atol = 1e-4,
+        lav_groups = Dict(:Pasteur => 1, :Grant_White => 2))
+end
+
+@testset "fitmeasures/se_ls" begin
+    solution_ls = sem_fit(model_ls_multigroup)
+    @test all(test_fitmeasures(
+        fit_measures(solution_ls), 
+        solution_lav[:fitmeasures_ls];
+        fitmeasure_names = fitmeasure_names_ls, rtol = 1e-2, atol = 1e-5))
+
+    update_partable!(
+        partable, identifier(model_ls_multigroup), se_hessian(solution_ls), :se)
+    @test compare_estimates(
+        partable, 
+        solution_lav[:parameter_estimates_ls]; atol = 1e-2,
+        col = :se, lav_col = :se,
+        lav_groups = Dict(:Pasteur => 1, :Grant_White => 2))
+end
+
+if !isnothing(specification_miss_g1)
+
+####################################################################
+# FIML estimation
+####################################################################
+
+model_g1 = Sem(
+    specification = specification_miss_g1,
+    observed = SemObsMissing,
+    loss = SemFIML,
+    data = dat_miss_g1,
+    imply = RAM,
+    diff = SemDiffEmpty()
+)
+
+model_g2 = Sem(
+    specification = specification_miss_g2,
+    observed = SemObsMissing,
+    loss = SemFIML,
+    data = dat_miss_g2,
+    imply = RAM,
+    diff = SemDiffEmpty()
+)
+
+model_ml_multigroup = SemEnsemble(model_g1, model_g2; diff = semdiff)
+
+############################################################################
+### test gradients
+############################################################################
+
+start_test = [
+    fill(0.5, 6); 
+    fill(1.0, 9); 
+    0.05; 0.01; 0.01; 0.05; 0.01; 0.05; 
+    fill(0.01, 9);
+    fill(1.0, 9); 
+    0.05; 0.01; 0.01; 0.05; 0.01; 0.05;
+    fill(0.01, 9)]
+
+@testset "fiml_gradients_multigroup" begin
+    @test test_gradient(model_ml_multigroup, start_test; atol = 1e-7)
+end
+
+
+@testset "fiml_solution_multigroup" begin
+    solution = sem_fit(model_ml_multigroup)
+    update_estimate!(partable_miss, solution)
+    @test compare_estimates(
+        partable_miss, 
+        solution_lav[:parameter_estimates_fiml]; atol = 1e-4,
+        lav_groups = Dict(:Pasteur => 1, :Grant_White => 2))
+end
+
+@testset "fitmeasures/se_fiml" begin
+    solution = sem_fit(model_ml_multigroup)
+    @test all(test_fitmeasures(
+        fit_measures(solution), 
+        solution_lav[:fitmeasures_fiml]; rtol = 1e-3, atol = 0))
+
+    update_partable!(
+        partable_miss, identifier(model_ml_multigroup), se_hessian(solution), :se)
+    @test compare_estimates(
+        partable_miss, 
+        solution_lav[:parameter_estimates_fiml]; atol = 1e-3,
+        col = :se, lav_col = :se,
+        lav_groups = Dict(:Pasteur => 1, :Grant_White => 2))
+end
+
+end