When suppressor effects may reflect multicollinearity artifacts: a cautionary tale from multiple object tracking and...

To evaluate whether multiple object tracking (MOT) predicts clay pigeon shooting performance, and to illustrate how multicollinearity between demographic covariates can mimic suppressor effects.

Twenty shooters completed a MOT task (2-6 targets). Zero-order and partial correlations, hierarchical regression, LOOCV, VIF diagnostics, and power analysis were computed. Two specifications were compared: Specification A (age + training years, multicollinear) and Specification B (age only, VIF < 2.5).

Age and training years were near-perfectly correlated (r = 0.928). Under Specification A $r=0.928;\mathrm{VIF}=11.7$ , MOT-5 and MOT-6 appeared as nominally significant partial correlates (r_partial ≈ 0.49, p < 0.03, uncorrected), a classical suppression pattern. Under Specification B, these effects attenuated to non-significance (p > 0.13). The training coefficient reversed sign (B = -2.12 despite r = +0.65), consistent with multicollinearity distortion. Power analysis indicated N ≈ 68 would be required to detect the residual effect (f² = 0.15), though this estimate is illustrative given the small sample. $\mathrm{AUC}=0.660f2=\mathrm{0.12.}$

The apparent suppressor effect is consistent with multicollinearity between age and training years rather than genuine prediction, though the small sample (N = 20) precludes definitive conclusions. Routine VIF, bootstrap stability checks, and power analysis are recommended to avoid artifacts in small-sample sports cognition research.