opie
New Member
Posts: 1
|
Post by opie on Jun 23, 2014 19:28:32 GMT
Hi Jian,
Thank you for providing the useful software. In your paper of genomic partitioning of genetic variation for complex traits, you tested the estimates of genic and intergenic heritability against the expected values from the genic and intergenic coverages by a goodness of fit test. How did you calculate the p value?
Did you fit both genic and intergenic GRMs together in the mixed model, and then calculate the log likelihood for the full model (logL) and that for the reduced model (logL0)? Is the p-value reported in the *.hsq file for LRT = 2[logL - logL0] the p value that you used in your paper to test significance of one genomic region?
Thank you
|
|
|
Post by Jian Yang on Jun 25, 2014 6:52:45 GMT
re 1) In Yang et al. 2012 Nat Genet, we used goodness of fit test to test if genic regions explain disproportional more proportion of variance. en.wikipedia.org/wiki/Goodness_of_fitO = observed proportion E = expected proportion based on coverage re 2) This test is to test against 0 rather than an expected proportion based on coverage.
|
|
|
Post by MTL on Dec 24, 2014 0:32:58 GMT
Hi Jian,
I've read your papaer in Nat Genet (2012), and I also want to calculate the p value for the goodness of fit test of estimated genic heritability against the expected value. I tried your estimates to calculate chi-squared value. I treated 0.256/0.419 as the observed proportion, 0.358 as the expected proportion, and (0.023*0.023)/(0.419*0.419) as the variance. I then computed the value of square of (O-E) divided by the variance that follows chi-squared distribution. I don't know whether the above procedures are correct? Could you give me the instruction about this calculation?
Thank you
|
|