We used linear regression to evaluate correlation of gene-expression levels within local haplotype blocks. We measured effect size by using the regression slope, β, and the coefficient of determination, R2. The linear model we used considers additive effects of two haplotype blocks. More specifically, for each block, the two parental haplotypes of each child are encoded with two covariates, p and m. The maternal haplotype mi of child i, for example, is either 0 or 1, depending on which of the two possible maternal alleles is present. Then, an expression trait is regressed as the summation of effects of two parental haplotypes, Ti ∼ μ + βjpi + βkmi, where Ti is the trait of individual i, the effects of two parental alleles k and j are expressed by βj and βk, and μ is the intercept. Each sibling has two choices of parental haplotypes on each side—p,m∈{0,1}—to yield four total combinations. Gene expression Ti uses log2(FPKM [fragments per kilobase per million] values). For splicing quantifications, we used relative transcript abundances, which we calculated by dividing the FPKM of each isoform by the FPKM of the whole gene (see Table S5). For cis-eQTLs, we only tested the local haplotypes containing the genes, which is sufficient for including most cis-eQTL signals (Figures S3–S5). Furthermore, we confirmed gene-expression levels and eQTL effect sizes with existing microarray data on the same family (Figures S6 and S7).