ebic: Extended Bayesian Information Criterion

In countprop: Calculate Model-Based Metrics of Proportionality on Count-Based Compositional Data

Extended Bayesian Information Criterion

Description

Calculates the Extended Bayesian Information Criterion (EBIC) of a model. Used for model selection to asses the fit of the multinomial logit-Normal model which includes a graphical lasso penalty.

Usage

ebic(l, n, d, df, gamma)

Arguments

l	Log-likelihood estimates of the model
n	Number of rows of the data set for which the log-likelihood has been calculated
d	The size of the (k-1) by (k-1) covariance matrix of a k by k count-compositional data matrix
df	Degrees of freedom
gamma	A tuning parameter. Larger values means more penalization

Value

The value of the EBIC.

Note

The graphical lasso penalty is the sum of the absolute value of the elements of the covariance matrix Sigma. The penalization parameter lambda controls the sparsity of Sigma.

Examples

data(singlecell)
mle <- mleLR(singlecell, lambda.gl=0.5)
log.lik_1 <- mle$est[[1]]$log.lik
n <- NROW(singlecell)
k <- NCOL(singlecell)
df_1 <- mle$est[[1]]$df

ebic(log.lik_1, n, k, df_1, 0.1)

countprop documentation built on Aug. 18, 2023, 9:06 a.m.