Association analysis of rs1344706 was carried out using a likelihood procedure described previously.11 For the genome-wide typed samples, a surrogate for rs1344706 made up of a linear combination of two-marker haplotypes, defined using the HapMap CEU, was used. This method, which we have used earlier,12 is an extension of the two-marker haplotype tagging described in Pe’er et al.13 and is similar in spirit to the methods described in Nicolae14 and Zaitlen et al.15 Genomic control16 was used to correct for relatedness and potential population stratification in each genome-wide typed study group. With the exception of Iceland, genomic control factors were < 1.1; in Iceland, some related individuals were included in the analysis and genomic control factors were 1.19, 1.12 and 1.18 for the schizophrenia, bipolar and psychosis analyses, respectively. The study groups within the Illumina genome-wide typed and follow-up sets were combined using the Mantel–Haenszel model.17 The combined Illumina genome-wide typed and combined follow-up sets were joined using summary statistics. P-values were calculated by summing z-scores with each data set’s z-score multiplied by the inverse of that data set’s s.e.
