Table 4.

Profile associations with outcomes: naive vs. uncertainty-corrected estimation (N = 489).

Outcome	Naive modal-assignment test	BCH-corrected estimate (bootstrap 95% CI)
Workshop attendance	χ2(3) = 25.59, p < 0.001, V =0.23	ROBUST. Low 32.5% vs High 6.6%: +24.9 pp [+9.1, +38.8]; vs Average 10.2%: +20.5 pp [+1.4, +35.5]; vs Discordant 10.9%: +18.4 pp [−18.5, +41.7] (ns). Unrelated to department (p =0.107) and year (p =0.107).
Stress-management confidence	H(3) = 17.56, p < 0.001, ε2 =0.030	ROBUST. Discordant highest (3.54): vs Average 2.59, +0.90 [+0.44, +1.39]; High 3.13 vs Average: +0.48 [+0.08, +0.91]. Discordant vs High +0.42 [−0.12, +0.89] and Low 3.02 vs Average +0.43 [−0.03, +0.89] not excluding zero.
GPA band code	H(3) = 6.68, p =0.083, ε2 =0.008	No association. All pairwise CIs include zero; High EI descriptively highest (band-code mean 3.94 on the 1–5 scale).

BCH, bias-adjusted three-step; pp, percentage points; CIs from 1,471 valid bootstrap resamples (of 2,000; centroid matching). GPA values are means of ordered band codes (1 = ≤ 2.0 … 5 = ≥3.5), not actual grade point averages. Across all 18 pairwise contrasts, the Low-vs-High workshop, Discordant-vs-Average stress, and Low-vs-Average workshop contrasts survived Holm correction; the workshop and stress omnibus associations survived Bonferroni correction across the three outcome families.