Construct Relational Covariates from a Network Array

Builds standard relational covariates from a base network array: the original (main) effect, the reciprocal (transpose) effect, and a transitive closure effect. These are common exogenous covariates (Z) in SIR models capturing higher-order network dependencies.

Usage

rel_covar(arr, name, effects = c("main", "reciprocal", "transitive"))

Arguments

arr

A 3D array (m x m x T) of network data. Typically an outcome variable or a covariate that varies across dyads and time.

name

Character string used to label the covariates in the output array's third dimension. The main effect is labeled name, the reciprocal paste0(name, "_recip"), and the transitive paste0(name, "_trans").

effects

Character vector specifying which relational covariates to include. Any subset of c("main", "reciprocal", "transitive"). Default is all three.

Value

A 4D array (m x m x q x T) where q is the number of requested effects (1 to 3). Suitable for passing directly as the Z argument to sir.

Details

Main

The original array, Z_ij = arr_ij. This captures the direct dyadic effect.

Reciprocal

The transpose, Z_ij = arr_ji. Captures whether the reverse relationship matters.

Transitive

A measure of shared connectivity: Z_ij = (S network. Captures triadic closure and transitivity effects.

Examples

if (FALSE) { # \dontrun{
# Build relational covariates from trade data
Z_trade <- rel_covar(trade_array, "trade")
dim(Z_trade)  # m x m x 3 x T

# Use only main and reciprocal effects
Z_simple <- rel_covar(trade_array, "trade", effects = c("main", "reciprocal"))

# Pass to sir() as exogenous covariates
fit <- sir(Y, W, X, Z = Z_trade, family = "poisson")
} # }