nicotine. As the threshold was relaxed and more SNPs were included, improvement is seen for every phenotype (sans nicotine), until the effect plateaus at around .05. This pattern of results is generally true for each of the LD cutoffs (each of the three graphs in the top row of Figure 3). Most obviously, there appears to be a polygenetic effect—the variance accounted for in the phenotype increases substantially as the proportion of SNPs is increased. Results appear to be generally dampened by the choice of LD cutoff, although not substantially so. The polygenetic effect, and the pattern of results, is true for the phenotypic data as well as the simulated phenotypes, regardless of the number of SNPs contributing to the phenotype in the simulations. The random, non-genetically-related phenotype (bottom right figure in Figure 3) shows no association, as expected.
