BicZ: Compute the BIC of a given structure

In CorReg: Linear Regression Based on Linear Structure Between Variables

Description

Compute the BIC of a given structure

Usage

BicZ(X = X, Z = Z, Bic_null_vect = NULL, Bic_old = NULL, methode = 1,
  Zold = NULL, star = FALSE)

Arguments

X	the dataset
Z	binary adjacency matrix of the structure (size p)
Bic_null_vect	the BIC of the null hypothesis (used for independent variables)
Bic_old	BIC (vector) associated to Zold
methode	parameter for OLS (matrix inversion) methode_BIC parameter for OLS (matrix inversion) 1:householderQr, 2:colPivHouseholderQr
Zold	another structure with some common parts with Z (allows to compute only the differences, to be faster)
star	boolean defining wether classical BIC or BIC* (over-penalized by a hierarchical uniform assumption to avoid over-learning)is computed

Value

The vector of the BICs associated to each covariate (conditionnal distribution) according to the sub-regression structure.

Examples

   ## Not run: 
   require(CorReg)
data=mixture_generator(n=15,p=5,valid=0)#dataset generation
Z=data$Z #binary adjacency matrix that describes correlations within the dataset
X=data$X_appr
Bic_null_vect=density_estimation(X=X)$BIC_vect
#Computes the BIC associated to each covariate (optional, BicZ can do it if not given as an input)
#computes the BIC associated to the structure
res=BicZ(X = X,Z = Z,Bic_null_vect=Bic_null_vect)
    
## End(Not run)

CorReg documentation built on Sept. 6, 2019, 3 a.m.