chawn
New Member
Posts: 3 
|
Post by chawn on Jul 1, 2016 21:57:46 GMT
How was the P value calculated from --mlma option?
Running association tests for 58689 SNPs ...
It seems that parallel computing is not used for association test?
gvcta64 --mlma --bfile gfilled2 --grm grmRun --autosome-num 22 --autosome --pheno ff.phen --thread-num 14 --out gwas
Thanks!
Chawn
|
|
|
|
Post by Jian Yang on Jul 3, 2016 5:38:24 GMT
I've tested. It should be used. See my screenshot below.  |
|
chawn
New Member
Posts: 3
|
Post by chawn on Jul 5, 2016 14:03:08 GMT
Thank you very much for your response.
Is there any error in my command?
At the beginning the parallel is used, but after "Running association tests for 58689 SNPs ...", it seems only one cpu is used in my server.
Another question is that the phenotype (y) need to be standardized for running --reml? Thanks!
|
|
|
|
Post by Jian Yang on Jul 6, 2016 23:34:45 GMT
Re 1) maybe something to do with the setting of your cluster job.
Re 2) not have to.
|
|
chawn
New Member
Posts: 3
|
Post by chawn on Jul 13, 2016 17:05:53 GMT
Hi Yang,
I still did not get the solution why gcta64 use one cpu only after "Running association tests for 58689 SNPs ...". Do you have suggestions on how to check the setting of cluster?
When I run other jobs the cluster can do parallel well.
Thanks,
Chawn
|
|
|
|
Post by janeoeaf on Aug 8, 2016 17:17:49 GMT
Hi Yang
In GCTA, Is the p-value of snp test calculated from LRT? example: LRT= -2(logL0_model_without_candidate_snp - logL1_model_with_candidate_snp) LRT~chi-square(df=1)
Or other test, like wald test?
best regards
|
|
|
|
Post by Jian Yang on Aug 10, 2016 3:53:31 GMT
Yes, LRT = 2(logL1_model_with_g - logL0_model_without_g) ~ chi-squared(df = 1).
LRT ~= (Vg_estimate / SE)^2
|
|
|
|
Post by janeoeaf on Aug 10, 2016 14:15:02 GMT
Thanks Yang,
how about hypothesis test of candidate SNP (fixed effect)?
H0: SNPi==0
H1: SNPi!=0
How is made this hypothesis test?
best regards
|
|
|
|
Post by Jian Yang on Aug 15, 2016 12:08:48 GMT
If you fit the SNP genotype as a fixed covariate, you might use the --est-fix option. The estimate and SE will be reported in the log file.
|
|
