Node Crossing Vectors by Dr Keith Strang Reid Cailleach Computing Ltd 2026.
NXV is an exact combinatorial extension of Mann-Whitney U for tied data and small clinical rating scales.
This is BETA version 0.3b - "scores on the doors"
ENTER TWO SAMPLES and get a p-value. NXV allows for ties in a way that Mann-Whitney U does not. Perhaps you are using a rating scale. If you are just counting things, don't worry about r. But if you are using a scale that only has 7 possible scores, then put r as 7. If you are using a scale like Yes/Maybe/No that has three options, then put r as three. If you are scoring out of ten, put r as 10.
How many tied elements do you expect? That is λ. If you don't expect any ties then λ is 0. r and λ interact but the app will keep you right.
Parameters
Results
NXV null distribution will appear here
pNXV—
For comparison — the result a ties-blind analysis (classical
Mann-Whitney U, the untied null) would have reported on your
data. The NXV result above is the answer; this is the warning:
where the two differ, ignoring ties would have misled.
The dashed curve is the normal approximation to this untied
null (continuity-corrected), with its p shown in the graph
title.