compute_cAIC_for_model_set: Compute cAIC for different models

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

Description

Function compute_cAIC_for_model_set computes cAIC for different models in the model set

Usage

compute_cAIC_for_model_set(
  X,
  y,
  clusterID,
  X_cluster_full = NULL,
  model = "NERM",
  covariate_selection_matrix = NULL,
  modelset = "all_subsets",
  intercept = FALSE,
  common = NULL
)

Arguments

X	Matrix with covariates for fixed effects
y	Vector of responses
clusterID	Vector with cluster labels
X_cluster_full	Matrix with cluster level covariates for fixed effects of the full model. Default: NULL
model	Type of mixed model: NERM, FHM, RIRS (random slopes and random intercepts)
covariate_selection_matrix	Matrix composed of zeros and ones indicating fixed models in each parameter. Default: NULL
modelset	Type of model set: "all_subsets" - all possible models using all covariates "nested" - a sequence of nested models using all covariates "part_subset" - all possible models using a subset of covariates
intercept	Is column of ones, representing the intercept, present in X Default: TRUE
common	A vector indicating variables forced to be present in each model. Default: NULL

Details

• model - for the moment, only NERM is supported

• covariate_selection_matrix - necessary only if modelset = NULL

Value

List with parameters:

cAIC_min	Index of selected model
cAIC_models	cAIC for all considered parameters
degcAIC_models	Penalty for all considered models
beta_sel	Fixed effects of the selected model
mu_sel	Mixed effects of the selected model
indices_sel	Indices of the selected covariates among full covariate set
sig_u_sel	Variance parameter of random effects of the selected model
sig_e_sel	Variance parameter of errors of the selected model
X_full	Matrix with fixed effects covariates of the full model
X_cluster_full	Matrix with cluster level covariates for fixed effects of the full model
modelset_matrix	Matrix composed of zeros and ones. Ones correspond to covariate in a model which is represented in nth row

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)

cAIC_model_set =
compute_cAIC_for_model_set(X, y, clusterID,
                           model = "NERM",
                           covariate_selection_matrix = NULL,
                           modelset  = "part_subset",
                           common = c(1:8),
                           intercept = FALSE)

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