lgcp_IC: Information criteria for point processes

In inlabru: Bayesian Latent Gaussian Modelling using INLA and Extensions

Information criteria for point processes

Description

Calculate information criteria for point process models, using a fitted model and optionally new data. The interface is still under development and may change without warning in future versions.

Usage

lgcp_IC(fit, data, predictor, domain, samplers, n.samples = 2000L)

delta_IC(data)

Arguments

fit	A fitted bru model object, from bru() or lgcp().
data	A data frame containing the observed points, with columns for coordinates and any covariates used in the model. The data frame may also include a column named .block that indicates the block membership of each point, if blockwise criteria are to be calculated.
predictor	An expression that evaluates to the linear predictor of the model, given the data. This should be an expression that can be evaluated in the context of the data frame and model components, just as in the formula argument of bru_obs() and lgcp().
domain, samplers	Arguments defining the integration domain(s) and sampling regions, as for bru_obs() and lgcp().
n.samples	Integer. The number of posterior samples to draw for estimating the criteria. Default is 2000.

Value

A tibble with columns for the Criterion name, Method, log pointwise predictive density (lppd), effective number of parameters (p_eff), and the information criterion value (IC)

Functions

• delta_IC(): Calculate IC differences based on joined output from several lgcp_IC calls;

  delta_IC(rbind(cbind(lgcp_IC(...), Model = "A"),
                 cbind(lgcp_IC(...), Model = "B")))
  

inlabru documentation built on July 28, 2026, 9:07 a.m.