Partial correlations from a covariance matrix — partial

Inverts the covariance matrix (or takes the precision matrix as given) and standardizes its negative off-diagonal elements. Applied to the innovation covariance of the VAR(1), the result is the contemporaneous network: the partial correlations of the states at the same prompt given the previous prompt and the other states.

Usage

partial_cors(S, precision = FALSE)

Arguments

S: A covariance (or precision, with precision = TRUE) matrix.
precision: Logical; is S already a precision matrix?

Value

The matrix of partial correlations (the contemporaneous network when S is an innovation covariance) with unit diagonal.

Examples

round(partial_cors(default_params()$Psi), 2)
#>         NegA PosA Stress Fatigue
#> NegA     1.0 -0.3    0.3     0.0
#> PosA    -0.3  1.0    0.0     0.0
#> Stress   0.3  0.0    1.0     0.2
#> Fatigue  0.0  0.0    0.2     1.0
# from a precision matrix directly
K <- solve(default_params()$Psi)
all.equal(partial_cors(K, precision = TRUE), partial_cors(default_params()$Psi))
#> [1] TRUE