Cases satisfied criteria for SCZ. Clinical characteristics and copy number variation have been described previously15. DNA was extracted from whole blood, with approval from institutional review boards. Genotypes were called using the Birdseed/Birdsuite algorithm25 and analyses were performed with PLINK v1.0526. Association analyses used a Cochran-Mantel-Haenszel test and logistic regression with covariates for sample site and ancestry. In the simulations, we generated datasets with pairs of unobserved variants and marker SNPs in varying degrees of within-pair LD, based on the effective number of independent SNPs in the ISC and assuming Hardy-Weinberg equilibrium and linkage equilibrium between different pairs of SNPs. We considered a large grid of possible values for allele frequency and effect size distributions, also varying the proportion of non-null variants and the LD between causal allele and observed marker. We retained models that produced similar profiles of target sample R2 compared to the original ISC analysis, for the same range of pT thresholds, and calculated the implied total genetic variance under these models, assuming additivity within and across loci. See Supplementary Information for details.
