One-sample sign-flip max-T inference on transported subject maps

Computes cluster-wise one-sample t-statistics across subjects on transported values (SxQ matrix), and calibrates p-values by the max-|t| distribution under random subject-wise sign flips (symmetric null). This does not re-estimate DKGE, leveraging LOSO independence of each subject's value.

Usage

dkge_signflip_maxT(
  Y,
  B = 2000,
  center = c("mean", "median"),
  tail = c("two.sided", "greater", "less")
)

Arguments

Y

SxQ matrix of subject values on the medoid parcellation (rows=subjects, cols=clusters)

B

number of sign-flip permutations

center

"mean" or "median" for the location statistic (t uses mean)

tail

"two.sided" | "greater" | "less"

Value

list with fields: stat (Q-vector), p (Q-vector), maxnull (B-vector), flips (SxB signs)