TABLE 5.

Genetic correlation estimates between population groups (POP1, POP2, and POP3) by bivariate GREML for two phenotypes.

Phenotype	Genetic correlation between POP1 and POP2			Genetic correlation between POP2 and POP3			Genetic correlation between POP1 and POP3
	Estimate	SE	P-value	Estimate	SE	P-value	Estimate	SE	P-value
Qualifications	0.2554	0.2223	8.09E−04	0.4795	0.1550	7.85E−04	0.5676	0.2743	0.1149
Age first had sexual intercourse	0.7418	0.3984	0.5169	0.0491	0.2284	3.14E−05	1.2176	0.3629	0.5488

The phenotypes were adjusted by basic plus additional confounders of fixed effects and transformed by rank-based INT. The bivariate GREML results for qualifications indicated a significant genetic heterogeneity between POP1 and POP2 (p-value = 8.09E−04), and between POP2 and POP3 (p-value = 7.85E−04), but showed no genetic heterogeneity between POP1 and POP3. These results were consistent with our findings from the G × P RNM. For age first had sexual intercourse, the bivariate GREML detected a significant heterogeneity between POP2 and POP3 (p-value = 3.14E−05), however, there was no interaction signal between POP1 and POP3 (as expected). Unexpectedly, the bivariate GREML failed to find genetic heterogeneity across POP1 + POP2 although RNM provided a significant signal. SE denotes standard error. P-value was obtained through a Wald test under a null hypothesis that genetic correlation equals to 1.