|
Post by mengze1 on Jan 8, 2021 4:07:13 GMT
I am trying to calculate heritabilities using mixed linear model. I want to test the significance of single variance component. For example, I partitioned the SNPs as Chromosome. There will be 22 chromsomes and 1 kinship matrix. I wanna test if a single variance component is significant. My command is: gcta64 --mgrm file.txt --pheno file.phen --reml --reml-alg 2 --reml-maxit 10000 --reml-lrt 1 --out output
The result shows
Sum of V(G)/Vp 0.293172 0.080451
logL 184.941
logL0 184.948
LRT 0.000
df 1
Pval 5.0000e-01
n 407
I wonder if this is legit? How reduced model's loglikelihood is larger than the full model.
|
|
|
Post by codecat on Jul 16, 2021 2:58:36 GMT
Hi mengze1,
I hope you have solved your question by now. May I ask your a simple question about the kinship matrix? How did you apply the kinship matrix in the gcta64 command?
Thank you!
|
|