Detection of Fetomaternal Genotype Associations in Early-Onset Disorders: Evaluation of Different Methods and Their Application to Childhood Leukemia

Figure 4

Type I error rate for the HD and HD-NPC approaches for the maternal association test assuming either mating symmetry or asymmetry. Type I error rates are shown as a function of departure from mating symmetry, as measured by Δ𝐶, under the null model where GC₁₁ = GC₁₂ = GC₂₂ = GM₁₁ = GM₁₂ = GM₂₂ = 1 (Table 2, Model 5). Allele prevalence is set at 𝑞=0.3. Log-linear, likelihood-based testing (Mother versus Null) was performed assuming either mating asymmetry estimating nine mating-type parameters in the likelihood-ratio tests (Asym), or mating symmetry using only six mating-type parameters in the likelihood-ratio tests (Sym), or with no a priori hypothesis, using a 3 degree-of-freedom likelihood-ratio test to first test the assumption of mating symmetry (Mating Asymmetry (9df) versus Mating Symmetry (6df)) and then using either the mating asymmetry models when asymmetry was detected or the mating symmetry models when it was not (Mixed). Horizontal reference line indicates type I error rate of 𝛼 = 0.05. HD, hybrid design using parents of controls in a log-linear framework; HD-NPC, log-linear analysis using unrelated controls rather than their parents.