Post by mckeller on Apr 22, 2016 21:54:27 GMT
Hi Jian,
Two quick questions:
1) We would like to test whether the overall h2_snp increases as a function of a moderating environmental variable (really, what could be called a heritability-be-environment interaction, or what is called a "quantitative" g*e interaction in twin models). The g*e analysis in gcta doesn't really get at this question (it tests whether the gene effects are correlated on average across the levels of the environmental variables - what is called a "qualitative" interaction in twin models). So for example, if males have higher heritability than females but the same genes affect the trait in similar ways in both sexes, that would be quantitative interaction. If gene effects differ across sexes but the heritabilities are the same, that would be qualitative interaction.
We're trying to understand whether there is a way to test for a quantitative (heritability-by-e) interaction in GCTA without resorting to splitting our sample and losing power. Are you aware of a way to do this? If not, we're considering trying to develop a way to do it using HE-regression.
2) Is there a way to use imputed dosages directly to estimate the GRM rather than hard-calls from imputed data? The former would seem to better account for uncertainty in imputed SNP calls.
Thanks!
Matt
Two quick questions:
1) We would like to test whether the overall h2_snp increases as a function of a moderating environmental variable (really, what could be called a heritability-be-environment interaction, or what is called a "quantitative" g*e interaction in twin models). The g*e analysis in gcta doesn't really get at this question (it tests whether the gene effects are correlated on average across the levels of the environmental variables - what is called a "qualitative" interaction in twin models). So for example, if males have higher heritability than females but the same genes affect the trait in similar ways in both sexes, that would be quantitative interaction. If gene effects differ across sexes but the heritabilities are the same, that would be qualitative interaction.
We're trying to understand whether there is a way to test for a quantitative (heritability-by-e) interaction in GCTA without resorting to splitting our sample and losing power. Are you aware of a way to do this? If not, we're considering trying to develop a way to do it using HE-regression.
2) Is there a way to use imputed dosages directly to estimate the GRM rather than hard-calls from imputed data? The former would seem to better account for uncertainty in imputed SNP calls.
Thanks!
Matt
