Hidden Cluster
Can 20 points change the story? Reveal a distant cluster and watch Pearson’s correlation respond. All 200 points are original, deliberately constructed examples—not observations of the real world.
The same axes, a different relationship
X and Y are unitless. Axes remain fixed when groups change. Blue circles = A; orange squares = B; purple triangles = C.
Read the change, then question it
Pearson’s r measures linear association, not causation. A distant group can alter the overall value even when most points stay unchanged. Try hiding A or B to see why the choice of included observations matters.
This dataset was tuned to a 0.120 absolute change between A+B and A+B+C. That is an illustration of sensitivity, not an estimate of a typical effect. Empty, single-point and zero-variance selections have no defined Pearson r.