Skip to contents

Convert precision matrix to partial correlation matrix

precision_to_pcor.Rd

Given a precision matrix \(\Omega\), returns the partial correlation matrix \(P\) with entries \(P_{ij} = -\Omega_{ij}/\sqrt{\Omega_{ii}\,\Omega_{jj}}\) and ones on the diagonal. Partial correlations measure the linear association between variables i and j after conditioning on all other variables.

Usage

precision_to_pcor(Omega)

Arguments

Omega

A square numeric matrix (precision / inverse covariance). Must have positive diagonal entries.

Value

A symmetric numeric matrix of the same dimension as Omega, with diagonal entries equal to 1 and off-diagonal entries in [-1, 1].

See also

posterior_pcor(), fitted.multiggm_fit()

Examples

Omega <- matrix(c(2, -0.5, -0.5, 1.5), 2, 2)
precision_to_pcor(Omega)
#>           [,1]      [,2]
#> [1,] 1.0000000 0.2886751
#> [2,] 0.2886751 1.0000000