chatterjee_matrix: Pairwise Chatterjee's xi correlation matrix
Description
Computes the p x p asymmetric matrix of Chatterjee's xi correlations
between all item pairs. Entry [j, k] is xi(item_j, item_k), which in
general differs from xi(item_k, item_j); this asymmetry is what
enables direction detection in subsequent graph-construction steps.
Each off-diagonal entry is computed by xi_stable() to average
over tie-breaking.
Usage
chatterjee_matrix(
U,
na = NULL,
Z = NULL,
w = NULL,
B = 1000,
seed = NULL,
verbose = FALSE
)
Value
A p x p numeric matrix with item labels as row/column names.
The diagonal is 1; off-diagonal entries are asymmetric.
Arguments
U
Either an exametrika object or raw data.
na
Values to be treated as missing.
Z
Missing indicator matrix.
w
Item weight vector.
B
Number of bootstrap replications per pair. Default 1000.
seed
Optional integer seed for reproducibility.
verbose
Logical. If TRUE, progress messages are displayed.
Details
Pairs are computed with pairwise-complete observations: for each
pair (j, k), rows where either Q[i, j] or Q[i, k] is missing are
excluded. The diagonal is set to 1.
References
Chatterjee, S. (2021). A new coefficient of correlation. Journal of
the American Statistical Association, 116(536), 2009-2022.