Documentation

Composite type to represent a dyadic kernel density estimator.

DyadicKernelDensityEstimator(kernel_name, bandwidth, significance_level,
                             n_resample, sdp_solver, evals, W, meta)

Construct a dyadic kernel density estimator.

Arguments

• kernel_name::String: which kernel to use.
• bandwidth::Float64: the bandwidth for the estimator.
• significance_level::Float64: for the confidence band/intervals.
• n_resample::Int: the number of resamples used to construct the confidence band/intervals.
• sdp_solver::String: semi-definite program solver.
• evals::Vector{Float64}: points at which to evaluate the density estimator.
• W::UpperTriangular{Float64}: array of dyadic data.
• meta::Dict: any extra information to pass to the estimator.

display(est::DyadicKernelDensityEstimator)

Display a dyadic kernel density estimator.

estimate_ROT_bandwidth(W::UpperTriangular{Float64},
                       kernel_name::String)

Estimate a rule-of-thumb bandwidth from dyadic data.

fit(est::DyadicKernelDensityEstimator)

Fit a dyadic kernel density estimator to data.

Composite type to represent a counterfactual dyadic kernel density estimator.

CounterfactualDyadicKernelDensityEstimator(kernel_name, bandwidth, significance_level,
                                           n_resample, sdp_solver, evals,
                                           W, X0, X1, meta)

Construct a counterfactual dyadic kernel density estimator.

Arguments

• kernel_name::String: which kernel to use.
• bandwidth::Float64: the bandwidth for the estimator.
• significance_level::Float64: for the confidence band/intervals.
• n_resample::Int: the number of resamples used to construct the confidence band/intervals.
• sdp_solver::String: semi-definite program solver.
• evals::Vector{Float64}: points at which to evaluate the density estimator.
• W::UpperTriangular{Float64}: array of treated dyadic data.
• X0::Vector{Int}: categorical vector of untreated covariates.
• X1::Vector{Int}: categorical vector of treated covariates.
• meta::Dict: any extra information to pass to the estimator.

Composite type to represent a counterfactual dyadic kernel density estimator.

ParametricCounterfactualDyadicKernelDensityEstimator(kernel_name, bandwidth, significance_level,
                                                     n_resample, sdp_solver, evals,
                                                     W, X0, X1, meta)

Construct a parametric counterfactual dyadic kernel density estimator.

Arguments

• kernel_name::String: which kernel to use.
• bandwidth::Float64: the bandwidth for the estimator.
• significance_level::Float64: for the confidence band/intervals.
• n_resample::Int: the number of resamples used to construct the confidence band/intervals.
• sdp_solver::String: semi-definite program solver.
• evals::Vector{Float64}: points at which to evaluate the density estimator.
• W::UpperTriangular{Float64}: array of treated dyadic data.
• X0::Vector{Int}: categorical vector of untreated covariates.
• X1::Vector{Int}: categorical vector of treated covariates.
• phat0::Vector{Float64}: vector of precomputed parametric untreated covariate estimates.
• phat1::Vector{Float64}: vector of precomputed parametric treated covariate estimates.
• meta::Dict: any extra information to pass to the estimator.

display(est::CounterfactualDyadicKernelDensityEstimator)

Display a counterfactual dyadic kernel density estimator.

display(est::ParametricCounterfactualDyadicKernelDensityEstimator)

Display a parametric counterfactual dyadic kernel density estimator.

fit(est::CounterfactualDyadicKernelDensityEstimator)

Fit a counterfactual dyadic kernel density estimator to data.

fit(est::ParametricCounterfactualDyadicKernelDensityEstimator)

Fit a parametric counterfactual dyadic kernel density estimator to data.

get_RIMSE(fhat::Vector{Float64}, f::Vector{Float64})

Compute the root integrated mean squared error of an estimate for a function.

get_average_width(cb::Matrix{Float64})

Return the average width of a confidence band.

get_coverage(cb::Matrix{Float64}, f::Vector{Float64})

Check if a confidence band covers the true density function.

get_f(p::Vector{Float64}, evals::Vector{Float64})

Get the true density function from example dyadic Gaussian mixture data.

make_dyadic_data(n_data::Int, p::Vector{Float64})

Generate some example dyadic data using a Gaussian mixture model.

mean(x::Vector{<:Number})

Compute the mean of a vector of numbers.

phi(t::Real)

Compute the standard normal density function.