|
Post by asling on May 14, 2016 22:40:33 GMT
Hello,
I am using GCTA bivariate GREML analysis to estimate the genetic correlation of two binary traits. Multiple GRMs has been fitted simultaneously in the model and the correlation of variance component V1 between the two traits has been tested for a null hypothesis of r=0 using --reml-bivar-lrt-rg 0.
Below is the result in .hsq file: The estimate of variance explained on the observed scale is transformed to that on the underlying scale: (Proportion of cases in the sample = 0.096971; User-specified disease prevalence = 0.090000 for disease 1 and = 0.130000 for disease 2) V(G1)/Vp_tr1_L 0.216717 0.073060 V(G1)/Vp_tr2_L 0.146261 0.066891 V(G2)/Vp_tr1_L 0.522873 0.073928 V(G2)/Vp_tr2_L 0.376051 0.062458 V(G3)/Vp_tr1_L 0.240340 0.058297 V(G3)/Vp_tr2_L 0.097680 0.048199 rG1 0.528551 0.215528 rG2 0.579340 0.081454 rG3 1.000000 0.207978 logL 27787.546 logL0 27786.371 (when rG fixed at 0.000) LRT 2.350 df 1 Pval 0.06265 (one-tailed test)
I am a bit confused about the Pvalue of rG1. As from the estimate and se, rG1=0.53>2*se=2*0.21, suggesting a significance at 5%, however from the Pvalue is the estimate of rG1 is within the 95%. May I ask if anyone has any idea about the interpretation of this result?
Thanks, Best wishes,
Asling
|
|