To further explore the impact of negative selection, we sampled β from a multi-normal distribution with mean 0 and variance \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$2p_{jl}\left( {1 - p_{jl}} \right)^S\sigma _\beta ^2$$\end{document}2pjl1−pjlSσβ2, where \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\sigma _\beta ^2$$\end{document}σβ2 is the variance of causal effect sizes. We considered three scenarios corresponding to (i) equal strength of selection in both ancestries (S1 = S2 = −0.5), (ii) stronger selection in Population 1 (S1 = −0.75 and S2 = −0.5) and (iii) stronger selection in Population 2 (S1 = −0.5 and S2 = −0.75). The phenotypes were generated in the same way as described above. For simplicity, we focused on a trait with a heritability h2 = 0.5 and controlled by MC = 5000 causal variants.
