The effect of these two types of missingness was studied in three genotype-phenotype models: 1) 1-factor Rasch model with the GV effect on the factor (Figure 1a; Tables S19, S20), 2) 1-factor Rasch model with the GV effect specific to one phenotype (Figure 1e; specific phenotype not showing blockwise missingness; Tables S21, S22, or showing blockwise missingness; Table S23), 3) network model with the GV effect specific to one phenotype (Figure 1f; specific phenotype not showing blockwise missingness; Tables S24, S25, or showing blockwise missingness; Table S26). In all these models, the 20 phenotypes correlated .56 (power results including missingness can thus be compared to power results concerning the same models without missingness presented in Tables S2, S4 and S7). Note that equal correlations between all phenotypes represents the ideal situation in which all phenotypes are equally reliable, i.e., the effect of the missingness only depends on the pattern of missingness, not e.g. on the reliability of the individual phenotypes.
