compute_information_criteria: Compute information criteria

In KatarzynaReluga/postcAIC: Post-cAIC selection inference under Linear Mixed Models

Description

Function compute_infromation_criteria provides mAIC, cAIC and mBIC for NERM

Usage

compute_information_criteria(
  X,
  y,
  clusterID,
  model,
  sig_u,
  sig_e,
  fit_model_fixed,
  fit_model_mixed
)

Arguments

X	Matrix with covariates for fixed effects
y	Vector of responses
clusterID	Vector with cluster labels
model	Type of mixed model: NERM, FHM, RIRS (random slopes and random intercepts)
sig_u	Variance parameter of random effects
sig_e	Variance parameter of errors
fit_model_fixed	Estimated model using fixed effects
fit_model_mixed	Estimated model using fixed and random effects

Details

Penalty term in cAIC depends on the model selected. Function compute_information_criteria is simplified because for now only NERM is supported.

Value

List with information criteria:

mAIC	Marginal AIC
mBIC	Marginal BIC
cAIC	Conditional AIC
deg_cAIC	Penalty of conditional AIC

Examples

n = 10
m_i = 5
m_total = 50

clusterID = rep(1:n, m_i)
p = 10
beta = rep(2, p)
u_i = rnorm(n, 0, 2)
u_i_aug = rep(u_i, each = m_i)
X = matrix(rnorm(m_total * p), m_total, p)
y = X%*%beta + u_i_aug + rnorm(m_total, 0, 1)
fit <- estimate_NERM(X, y, clusterID)

IC <- compute_information_criteria(X = X, y = y,
                                   clusterID = clusterID,
                                   model = "NERM",
                                   sig_u = fit$sig_u,
                                   sig_e = fit$sig_e,
                                   fit_model_fixed = fit$fit_model_fixed,
                                   fit_model_mixed = fit$fit_model_mixed)

KatarzynaReluga/postcAIC documentation built on Jan. 25, 2022, 12:33 a.m.