Selected-position counts cannot diagnose position bias unless arm exposure is balanced across positions. The diagnostic derives exposure first and refuses to emit a conclusion when it is not.

ARM × POSITION EXPOSURE

POS 1POS 2POS 3POS 4A8565B5793C7557D4749

EXPOSURE BALANCED: FALSE · POSITION EFFECT IDENTIFIED: FALSE

Arm by position exposure is unbalanced: B occupied position 3 nine times and position 4 three times.

V · 24 judgment occasions; each arm shown once per occasion · stage1.position.exposure_imbalance

No position effect is identified: selected-position marginals conflate arm preference with exposure.

Supersedes the withdrawn naive chi-square. An assignment-aware diagnostic is pending.

D · 24 judgment occasions · stage1.position.effect_not_identified