Covariance Selection with Bisection Optimization

Selects an optimal covariance matrix through BIC convergence.

Usage

covarianceSelectionBisection(
  S,
  rankedEdges,
  numberObservations,
  lowerBoundEdge,
  upperBoundEdge
)

Arguments

S

Required. A symetric p-by-p covariance matrix.

rankedEdges

Required. A list of ranked edges to be constrained by zero.

numberObservations

Required. Number of observations used to calculate BIC estimates.

lowerBoundEdge

Required. Numeric specifying the lower bound number of parameters (d) in BIC calculation: `BIC = -2 * loglikelihood + d * log(N)`

upperBoundEdge

Required. Numeric specifying the upper bound number of parameters (d) in BIC calculation: `BIC = -2 * loglikelihood + d * log(N)`

Value

A list containg

• `w` Estimated inverse covariance matrix.

• `resMiddle` The converged sparse inverse covariance matrix.

• `bicMiddle` The converged BIC estimate.

• `middleEdge` The converged estimate of parameters.