Create a Custom Covariance Structure for use with because

This helper function allows users to easily create custom covariance structures by providing just a function that computes the precision (or covariance) matrix. The function handles all the S3 method registration automatically.

Usage

because_structure(name, precision_fn, description = NULL)

Arguments

name: Character string; the name of your structure (e.g., "spatial_knn"). This will be used to create the S3 class and methods.
precision_fn: A function that takes your structure data and returns a precision matrix (N x N). The first return value should be the precision matrix. Additional list elements can be returned and will be passed to JAGS as data.
description: Optional description of the structure for documentation.

Value

A constructor function that creates structure objects of your custom class.

Details

This function creates:

A constructor function (returned) for creating structure objects
S3 methods for jags_structure_definition (generates JAGS code)
S3 methods for prepare_structure_data (prepares data for JAGS)

The precision function should return either:

A single matrix (the precision matrix)
A list with Prec (the precision matrix) and optionally other data

Examples

if (FALSE) { # \dontrun{
# Create a spatial distance-decay structure
spatial_decay <- because_structure(
  name = "spatial_decay",
  precision_fn = function(coords, decay_rate = 0.1) {
    dist_mat <- as.matrix(dist(coords))
    W <- exp(-decay_rate * dist_mat)
    diag(W) <- 0
    D <- diag(rowSums(W))
    D - 0.99 * W  # Precision matrix
  }
)

# Use it
my_struct <- spatial_decay(coords = my_coords, decay_rate = 0.2)
fit <- because(Y ~ X, data = data, structure = my_struct)
} # }