Figure 4From: A comparison of internal validation techniques for multifactor dimensionality reductionConservative (A) and liberal power (B) for epistatic models with two loci. Conservative and liberal power is plotted for increasing effect size for both MDR with a three-way split (3WS) and cross-validation (CV). Effect size is ordered first in terms of heritability and within a single level of heritability then in terms of odds ratio. Maximum standard errors for power estimates are 0.050 for both conservative and liberal power.Back to article page