Create a spatial random field model

Constructs a spatial random field model for MCMC inference. The spatial component can be either a mixture of directed graphical models (MDGM) or a Markov random field (MRF), specified via the spatial argument using the mdgm() or mrf() configuration helpers.

Usage

srf_model(nug, spatial = mdgm(), emission = NULL, n_colors = 2L)

Arguments

nug: A NaturalUndirectedGraph object defining the spatial structure.
spatial: A spatial configuration object created by mdgm() or mrf().
emission: Optional emission family for hierarchical models: "bernoulli", "gaussian", or "poisson". If NULL (default), creates a standalone model where the spatial field is observed directly.
n_colors: Number of categories for the spatial field (default 2).

Value

An SrfModel object.

Examples

edges <- rbind(c(1, 2), c(2, 1), c(2, 3), c(3, 2), c(1, 3), c(3, 1))
nug <- nug_from_edge_list(3, edges, seed = 42L)

# MDGM model
m1 <- srf_model(nug, spatial = mdgm(dag_type = "spanning_tree"))

# MRF model with exchange algorithm
m2 <- srf_model(nug, spatial = mrf(method = "exchange"),
                emission = "bernoulli")