Convert precision matrix to partial correlations

Computes partial correlations from a precision matrix using the formula pcor[i,j] = -Theta[i,j] / sqrt(Theta[i,i] * Theta[j,j]). Diagonal entries are set to 1.

Usage

precision_to_pcor(Theta)

Arguments

Theta

Precision matrix (p x p). Must be positive definite.

Value

A p x p partial correlation matrix with 1s on the diagonal. Returns a matrix of NAs if the diagonal contains non-positive or non-finite values.

Examples

Theta <- matrix(c(2, -0.5, -0.5, 2), 2, 2)
precision_to_pcor(Theta)  # partial correlation = 0.25
#>      [,1] [,2]
#> [1,] 1.00 0.25
#> [2,] 0.25 1.00