Table 2. Simulation - Genome-wide association study (GWAS).

The average proportion of datasets (±SE) in which the five true biomarkers are ranked among the top five according to two measures of variable importance, namely (1) original RSF algorithm (p_RSF), and (2) variable importance from the modified RSF algorithm (p₁). 1 − S_J+1, φ₁, φ₀ denote the cumulative incidence in the reference group, sensitivity and specificity, respectively. The results are stratified by Minor Allele Frequency categories of (0, 0.35] and (0.35, 0.5]. We assume the setting of no missed visits.

1 − Sj+1	φ1	φ0	No missing data
			MAF ∈ (0, 0.35]		MAF ∈ (0.35, 0.5]
			pRSF	p1	pRSF	p1
0.10	1.00	1.00	0.578(±0.0494)	0.560(±0.0496)	0.725(±0.0447)	0.701(±0.0458)
	0.75	1.00	0.575(±0.0494)	0.566(±0.0496)	0.702(±0.0457)	0.710(±0.0454)
	0.61	0.995	0.487(±0.0500)	0.511(±0.0500)	0.594(±0.0491)	0.657(±0.0475)
	1.00	0.90	0.339(±0.0473)	0.556(±0.0497)	0.435(±0.0496)	0.703(±0.0457)
0.30	1.00	1.00	0.619(±0.0486)	0.623(±0.0485)	0.737(±0.0440)	0.754(±0.0431)
	0.75	1.00	0.607(±0.0488)	0.578(±0.0494)	0.688(±0.0463)	0.692(±0.0462)
	0.61	0.995	0.529(±0.0499)	0.526(±0.0499)	0.571(±0.0495)	0.654(±0.0476)
	1.00	0.90	0.427(±0.0495)	0.563(±0.0496)	0.533(±0.0499)	0.767(±0.0423)