Skip to main content
. 2020 Dec 9;31(4):1873–1887. doi: 10.1093/cercor/bhaa327

Figure 2.

Figure 2

Identifying and correcting for the effects of subtle population stratification on GWAS results. (a) Correlations between SNP loadings on ancestry PCs (Beta_PCs) and GWAS effect sizes for full SA (Beta_Strat) demonstrate evidence for subtle population stratification (* indicates Bonferroni corrected significant correlation P-value < 0.0025, and o indicates a nominally significant correlation, P-value < 0.05). (b) Subtle population stratification is reduced after ancestry regression. (c) LD-score regression (LDSC) intercepts, standard measures of population stratification, are generally decreased after ancestry regression. An absence of population stratification and cryptic relatedness would be indicated by an LDSC intercept value of 1. (d) There is an inverse relationship between the degree of subtle population stratification (LDSC intercept prior to ancestry regression) and the amount of change caused by ancestry regression (cor[Beta_strat, Beta_r]). Error bars represent standard errors.