For the NTR and the QIMR-adult cohorts, the increase in statistical power for a GWAS on Neuroticism was determined that results from the increase in sample size and measurement precision due to the IRT test linking. A baseline condition of using 12 NEO-FFI items as in a previous meta-analysis (De Moor et al. 2010) was compared with using all available data from NEO-PI-R and other available inventories. We assumed that genotype data was non-missing for all phenotypes. Power was computed for a single nucleotide polymorphism (SNP) explaining 0.1 % of true phenotypic variance (latent trait) with allele frequency 0.5. Item data were simulated with parameter settings equal to the observed parameter estimates in the empirical data. Sample sizes were also the same as in the empirical data. For each power estimate, 100 data sets were simulated and analyzed, and the proportion of p-values smaller than 10−8 was calculated.
