jianyangqt / gcta

GCTA software
GNU General Public License v3.0
84 stars 26 forks source link

Question reg V(G) est in GCTA 1.94.1 #93

Closed RPorneso closed 2 months ago

RPorneso commented 2 months ago

Is the V(G) from the .hsq file the \sigma_g^2 in y = X\beta + g + e, with V = A\sigma_g^2 + I \sigma_e^2 in https://linkinghub.elsevier.com/retrieve/pii/S0002929710005987? Meaning it has to be multiplied by A to get the GBLUP per individual? My .hsq file is below where it shows LRT p val is almost 0.5. However, the residuals from .indi.blp are not almost equal to my phenotype. If I calculate it, the predicted value of the phenotype in my GBLUP is, on average, 0.26. I expected the V(G) in my .hsq file to be equivalent to A\sigma_g^2 and not sigma_g^2. Did I make a mistake somewhere in my script? Kindly advise.

.HSQ / .LOG

image image

.INDI.BLP

image

Thank you,

Ralph

RPorneso commented 2 months ago

the GRM I was provided is wrong. Closing this issue.

RPorneso commented 2 months ago

GRM diagonal entries are all almost 0