change coercion again...

TARGENE · May 29, 2024 · 53f985d · 53f985d
1 parent 2d6805a
commit 53f985d
Show file tree

Hide file tree

Showing 4 changed files with 14 additions and 10 deletions.
diff --git a/src/density_estimation/density_estimation.jl b/src/density_estimation/density_estimation.jl
@@ -45,13 +45,7 @@ function density_estimation(
     )
     outcome, parents = read_density_variables(density_file)
     dataset = TargetedEstimation.instantiate_dataset(dataset_file)
-    TargetedEstimation.coerce_types!(dataset, parents)
-    # Continuous and Counts except Binary outcomes are treated as continuous
-    if elscitype(dataset[!, outcome]) <: Infinite && !(TargetedEstimation.isbinary(outcome, dataset))
-        TargetedEstimation.coerce_types!(dataset, [outcome], rules=:discrete_to_continuous)
-    else
-        TargetedEstimation.coerce_types!(dataset, [outcome], rules=:few_to_finite)
-    end
+    coerce_parents_and_outcome!(dataset, parents, outcome=outcome)
 
     X, y = X_y(dataset, parents, outcome)
     density_estimators = get_density_estimators(mode, X, y)

diff --git a/src/estimation.jl b/src/estimation.jl
@@ -21,9 +21,6 @@ function estimate_from_simulated_data(
     origin_dataset = TargetedEstimation.instantiate_dataset(origin_dataset)
     sample_size = sample_size !== nothing ? sample_size : nrow(origin_dataset)
     estimands = TargetedEstimation.instantiate_estimands(estimands_config, origin_dataset)
-    for estimand in estimands
-        TargetedEstimation.coerce_types!(origin_dataset, estimand)
-    end
     estimators_spec = TargetedEstimation.instantiate_estimators(estimators_config)
     sampler = get_sampler(sampler_config, estimands)
     statistics = []

diff --git a/src/samplers/density_estimate_sampler.jl b/src/samplers/density_estimate_sampler.jl
@@ -46,6 +46,7 @@ function sample_from(sampler::DensityEstimateSampler, origin_dataset; n=100)
     sampled_dataset = sample_from(origin_dataset, sampler.all_parents_set; n=n)
 
     for (outcome, (parents, file)) in sampler.density_mapping
+        coerce_parents_and_outcome!(sampled_dataset, parents; outcome=nothing)
         conditional_density_estimate = Simulations.sieve_neural_net_density_estimator(file)
         sampled_dataset[!, outcome] = sample_from(
             conditional_density_estimate, 

diff --git a/src/utils.jl b/src/utils.jl
@@ -2,6 +2,18 @@
 ###                          Misc Functions                          ###
 ########################################################################
 
+function coerce_parents_and_outcome!(dataset, parents; outcome=nothing)
+    TargetedEstimation.coerce_types!(dataset, parents)
+    if outcome !== nothing
+        # Continuous and Counts except Binary outcomes are treated as continuous
+        if elscitype(dataset[!, outcome]) <: Infinite && !(TargetedEstimation.isbinary(outcome, dataset))
+            TargetedEstimation.coerce_types!(dataset, [outcome], rules=:discrete_to_continuous)
+        else
+            TargetedEstimation.coerce_types!(dataset, [outcome], rules=:few_to_finite)
+        end
+    end
+end
+
 function variables_from_dataset(dataset)
     confounders = Set([])
     outcome_extra_covariates = Set(["Genetic-Sex", "Age-Assessment"])