Computes the classical two-sample Hotelling \(T^2\) statistic comparing the
means of the rows of X assigned to cluster k1 versus cluster k2, using
a pooled (ridge-stabilized) covariance estimate.
Usage
hotelling_T2(X, cl, k1, k2, ridge = 1e-08)
Arguments
- X
A numeric data matrix with observations in rows.
- cl
Integer/character vector of cluster labels, one per row of X.
- k1, k2
The two cluster labels to compare.
- ridge
Small non-negative ridge added to the diagonal of the pooled
covariance for numerical stability.
Value
A single numeric \(T^2\) value, or NA_real_ if either cluster has
fewer than two members.
Examples
set.seed(1)
X <- matrix(rnorm(40), 20, 2)
cl <- rep(1:2, each = 10)
hotelling_T2(X, cl, 1, 2)
#> [1] 0.6052868
