|
Post by caramado on Sept 28, 2016 15:04:46 GMT
I'm running REML analyses with two matrices, a GRM that I computed using GCTA and a similarity matrix based on the environment that I created myself. GCTA runs normally but the errors for my customised matrix are extrangely large:
Source Variance SE
V(G1) 22.506469 0.329233
V(G2) 0.000040 1501060412147485935548689356428429334937600.000000
V(e) 16.555624 0.220699
Vp 39.062133 1501060412147485935548689356428429334937600.000000
V(G1)/Vp 0.576171 22140816387470270874146810279390317903872.000000
V(G2)/Vp 0.000001 38427468206399802086536012438265065897984.000000
Which obviously provides a large error as well in the heritabilities. The number of individuals is ~70,000 and I have used a similar matrix in the past (with fewer individuals) obtaining error values that looked normal (I was expecting a significant value of V(G2) ~5%, although this might be irrelevant). This happens for different traits, so I assume the problem is in my environmental matrix.
Do you have any idea of what might be causing this?
Thanks!!
|
|