Fixed typos and added docstrings

Author: dscolby

docs/make.jl

@@ -16,7 +16,7 @@ makedocs(;
     ),
     pages=[
         "CausalELM" => "index.md",
-        "Getting Started" => Any[
+        "User Guide" => Any[
             "Interrupted Time Series Estimation" => "guide/its.md",
             "G-computation" => "guide/gcomputation.md",
             "Double Machine Learning" => "guide/doublemachinelearning.md",

docs/src/reference/crossval.md

@@ -5,8 +5,8 @@ Methods to find the optimal number of neurons via cross validation
 CausalELM.CrossValidation
 CausalELM.CrossValidation.recode
 CausalELM.CrossValidation.generate_folds
-CausalELM.CrossValidate.validate_fold
-CausalELM.CrossValidate.cross_validate
-CausalELM.CrossValidate.best_size
-CausalELM.CrossValidate.shuffle_date
+CausalELM.CrossValidation.validate_fold
+CausalELM.CrossValidation.cross_validate
+CausalELM.CrossValidation.best_size
+CausalELM.CrossValidation.shuffle_data
 ```

docs/src/reference/estimation.md

@@ -8,7 +8,7 @@ CausalELM.Estimators.CausalEstimator
 CausalELM.Estimators.InterruptedTimeSeries
 CausalELM.Estimators.GComputation
 CausalELM.Estimators.DoubleMachineLearning
-CausalELM.Estimators.estimatecausaleffect!
+CausalELM.Estimators.estimate_causal_effect!
 CausalELM.Estimators.first_stage!
 CausalELM.Estimators.ate!
 CausalELM.Estimators.predict_propensity_score

docs/src/reference/inference.md

@@ -4,6 +4,6 @@ Methods for summarization and inference of estimators in the CausalELM package
 ```@docs
 CausalELM.Inference
 CausalELM.Inference.summarize
-CausalELM.Inference.quantitiesofinterest
-CausalELM.Inference.generatenulldistribution
+CausalELM.Inference.quantities_of_interest
+CausalELM.Inference.generate_null_distribution
 ```

src/inference.jl

@@ -321,7 +321,7 @@ function quantities_of_interest(m::Union{CausalEstimator, Metalearner}, n::Integ
 end
 
 """
-    quantities_of_interest(model, nsplits)
+    quantities_of_interest(model, n)
 
 Generate a p-value and standard error through randomization inference
 
@@ -345,16 +345,16 @@ julia> quantities_of_interest(its, 10)
 (0.0, 0.07703275541001667)
 ```
 """
-function quantities_of_interest(model::InterruptedTimeSeries, nsplits::Integer=1000, 
+function quantities_of_interest(model::InterruptedTimeSeries, n::Integer=1000, 
     mean_effect::Bool=true)
-    local null_dist = generate_null_distribution(model, nsplits, mean_effect)
+    local null_dist = generate_null_distribution(model, n, mean_effect)
     local metric = ifelse(mean_effect, mean, sum)
     local effect = metric(model.Δ)
 
     extremes = length(null_dist[effect .< abs.(null_dist)])
-    pvalue = extremes/nsplits
+    pvalue = extremes/n
 
-    stderr = (sum([(effect .- x)^2 for x in null_dist])/(nsplits-1))/sqrt(nsplits)
+    stderr = (sum([(effect .- x)^2 for x in null_dist])/(n-1))/sqrt(n)
 
     return pvalue, stderr
 end

src/metalearners.jl

@@ -405,6 +405,19 @@ julia> estimatecausaleffect!(m1)
 """
 estimate_causal_effect!(m::Metalearner) = estimate_causal_effect!(m)
 
+"""
+stage1!(x)
+
+Estimate the first stage models for an X-learner.
+
+This method should not be called by the user.
+
+```julia-repl
+julia> X, Y, T =  rand(100, 5), rand(100), [rand()<0.4 for i in 1:100]
+julia> m1 = XLearner(X, Y, T)
+julia> stage1!(m1)
+```
+"""
 function stage1!(x::XLearner)
     if x.regularized
         x.g = RegularizedExtremeLearner(x.X, x.T, x.num_neurons, x.activation)
@@ -423,9 +436,23 @@ function stage1!(x::XLearner)
     x.ps = predict(x.g, x.X)
 
     # Fit first stage outcome models
-    fit!(x.μ₀), fit!(x.μ₁)
+    fit!(x.μ₀); fit!(x.μ₁)
 end
 
+"""
+stage2!(x)
+
+Estimate the second stage models for an X-learner.
+
+This method should not be called by the user.
+
+```julia-repl
+julia> X, Y, T =  rand(100, 5), rand(100), [rand()<0.4 for i in 1:100]
+julia> m1 = XLearner(X, Y, T)
+julia> stage1!(m1)
+julia> stage2!(m1)
+```
+"""
 function stage2!(x::XLearner)
     d = ifelse(x.T === 0, predict(x.μ₁, x.X .- x.Y), x.Y .- predict(x.μ₀, x.X))
 

src/model_validation.jl

@@ -1,3 +1,8 @@
+"""
+Methods to test the sensitivity of interrupted time series estimators, G-computation, double 
+machine learning, S-learners, T-learners, and X-learners to violations of their modeling 
+assumptions.
+"""
 module ModelValidation
 
 using ..Estimators: InterruptedTimeSeries, estimate_causal_effect!, GComputation, 

src/models.jl

@@ -300,6 +300,24 @@ function placebo_test(model::ExtremeLearningMachine)
     return predict(model, model.X), model.counterfactual
 end
 
+"""
+    ridge_constant(model)
+
+Calculate the L2 penalty for a regularized extreme learning machine.
+
+For more information see: 
+    Li, Guoqiang, and Peifeng Niu. "An enhanced extreme learning machine based on ridge 
+    regression for regression." Neural Computing and Applications 22, no. 3 (2013): 
+    803-810.
+
+Examples
+```julia-repl
+julia> m1 = RegularizedExtremeLearner(x, y, 10, σ)
+ Extreme Learning Machine with 10 hidden neurons
+ julia> ridge_constant(m1)
+ 0.26789338524662887
+ ```
+ """
 function ridge_constant(model::RegularizedExtremeLearner)
     β0 = @fastmath pinv(model.H) * model.Y
     σ̃  = @fastmath ((transpose(model.Y .- (model.H * β0)) * (model.Y .- (model.H * β0))) / 
@@ -308,6 +326,23 @@ function ridge_constant(model::RegularizedExtremeLearner)
     return @fastmath first((model.H[2]*σ̃)/(transpose(β0)*transpose(model.H)*model.H*β0))
 end
 
+"""
+    set_weights_biases(model)
+
+Calculate the weights and biases for an extreme learning machine or regularized extreme 
+learning machine.
+
+For details see;
+    Huang, Guang-Bin, Qin-Yu Zhu, and Chee-Kheong Siew. "Extreme learning machine: theory 
+    and applications." Neurocomputing 70, no. 1-3 (2006): 489-501.
+
+Examples
+```julia-repl
+julia> m1 = RegularizedExtremeLearner(x, y, 10, σ)
+ Extreme Learning Machine with 10 hidden neurons
+ julia> set_weights_biases(m1)
+ ```
+ """
 function set_weights_biases(model::ExtremeLearningMachine)
     model.weights = rand(Float64, model.features, model.hidden_neurons)
     model.bias = rand(Float64, 1, model.hidden_neurons)