Skip to contents

Select Best Model Based on an Information Criterion

Source: R/best_model.R
best_BIC_model.Rd

Extracts the best-fitting model from a roams_SSM_list object according to a chosen information criterion, while excluding models with more than 50% outlying observations.

Usage

best_BIC_model(model_list, ic = c("BIC", "AIC", "HQIC"), k = NULL)

Arguments

model_list

An object of class roams_SSM_list.

ic

Character string specifying the information criterion. One of "BIC" (default), "AIC", or "HQIC" (Hannan-Quinn). Ignored if k is supplied.

k

Optional numeric penalty multiplier per model parameter, as in step. Overrides ic when non-NULL. Common values: 2 (AIC), log(n) (BIC), 2 * log(log(n)) (HQIC).

Value

A single roams_SSM object corresponding to the model with the smallest value of the chosen information criterion among those with fewer than 50% outlying observations.

Details

The criterion is computed as \(-2\ell + k \cdot p\), where \(\ell\) is the log-likelihood, \(p\) is the number of non-zero components, and \(k\) is the penalty multiplier. The ic argument provides named shortcuts; k allows a fully custom penalty and overrides ic when non-NULL.

Examples

if (FALSE) { # \dontrun{
# BIC (default)
best_BIC_model(models)

# Named shortcuts
best_BIC_model(models, ic = "AIC")
best_BIC_model(models, ic = "HQIC")

# Custom penalty via k (overrides ic)
best_BIC_model(models, k = 2)
} # }