Calculate Marginal Effects for Hierarchical Models — marginal

Computes the posterior distribution of average marginal effects by varying a target covariate over a grid.

Usage

marginal_effects(object, z_values, Z, burn = 0, verbose = TRUE, ...)

# S3 method for class 'rhierMnlRwMixture'
marginal_effects(object, z_values, Z, burn = 0, verbose = TRUE, ...)

# S3 method for class 'rhierLinearMixture'
marginal_effects(object, z_values, Z, burn = 0, verbose = TRUE, ...)

# S3 method for class 'rhierNegbinRw'
marginal_effects(object, z_values, Z, burn = 0, verbose = TRUE, ...)

Arguments

object: A fitted hierarchical model object (e.g., rhierMnlRwMixture, rhierLinearMixture, rhierNegbinRw).
z_values: A numeric matrix of grid values for the unit-level covariates Z. Each row defines one counterfactual Z*. Columns that are entirely NA are held at their training values; columns with no NAs are swept to the supplied grid value.
Z: A numeric matrix of unit-level covariates from the training sample (typically Data$Z from the original fit). Must have the same number of columns as z_values.
burn: Non-negative integer. Number of initial MCMC draws to drop before averaging (default 0).
verbose: Logical. Print progress per grid point (default TRUE).
...: Other arguments passed to methods.

Value

An object of class "marginal_effects"; see summary.marginal_effects() for downstream summarization.

Details

This is a generic function. Method implementations are provided for rhierMnlRwMixture, rhierLinearMixture, and rhierNegbinRw (the heter-cov subclasses inherit dispatch through their base classes).