
Scenario1
0.30/0.30 

Scenario2
0.30/0 

SNP 1 
SNP 2 
SNP 1 
SNP 2 
GBM 
1.3 (1.9) 
0.8 (1.2) 
1.1 (1.7) 
19.0 (19.9) 
RobustSNP 
3.0 (4.2) 
2.2 (3.3) 
1.1 (1.6) 
56.9 (33.7) 
AdditiveGWA 
1.7 (2.5) 
1.8 (2.7) 
0.8 (1.1) 
52.1 (35.3) 
Note: Results are presented as percentile median ranks of detecting SNP. A
robust measure of variability (Median Absolute Deviation, MAD) is given between
brackets. SNP1 always explains 0.3% of the variance, SNP2 either explains
0.3% (scenario1) or zero% (scenario2). Using GBM, a SNP with a zero main
effect (SNP2, scenario2) is within the19th percentile in 50% of the Monte Carlo
replications. Robust SNP and additive GWAS perform (as expected) only at
chance level (i.e., median percentile rank around 50). 