h2 estimate from genotype > h2 estimate from imputed data

h2 estimate from genotype > h2 estimate from imputed data Aug 7, 2023 0:51:49 GMT

Quote

Post by anbai on Aug 7, 2023 0:51:49 GMT

Hi, I ran GCTA to estimate the h2 for a quantitive trait using i) genotype arrays, and ii) imputed genotype data.

Here are the result logs:

Genotype arrays (tuto: yanglab.westlake.edu.cn/software/gcta/#GREMLanalysis):

Source Variance SE
V(G) 0.390146 0.016901
V(e) 0.608653 0.015923
Vp 0.998799 0.007107
V(G)/Vp 0.390615 0.016178
logL -20099.464
logL0 -20451.648
LRT 704.369
df 1
Pval 0.0000e+00
n 40952

Imputed data (tuto: yanglab.westlake.edu.cn/software/gcta/#GREMLinWGSorimputeddata):

Source Variance SE
V(G1) 0.097846 0.015257
V(G2) 0.071097 0.009219
V(G3) 0.066559 0.006642
V(G4) 0.044225 0.004166
V(e) 0.719333 0.015852
Vp 0.999060 0.007105
V(G1)/Vp 0.097938 0.015226
V(G2)/Vp 0.071164 0.009186
V(G3)/Vp 0.066621 0.006596
V(G4)/Vp 0.044267 0.004124

Sum of V(G)/Vp 0.279990 0.015710
logL -20136.583
logL0 -20158.071
LRT 42.976
df 1
Pval 2.7711e-11
n 40952

Can you explain why the former estimates a higher h2 than the latter?

Thanks