Does h2 from LDSC have similar problems with h2 from HE regression/GREML (e.g., Shared environmental effects, population stratification, MAF and LD heterogeneity)?
Replying to “Does h2 from LDSC have similar problems with h2 fr…”:
If the GWAS has issues then the LDSC h2 will have issues
Replying to “Does h2 from LDSC have similar problems with h2 fr…”:
but we cannot filer GRM or add PC to sum stats, etc. How do we correct for these?
Replying to “Does h2 from LDSC have similar problems with h2 fr…”:
they were corrected for in the GWAS
Replying to “Does h2 from LDSC have similar problems with h2 fr…”:
Finer points: Shared environment not relevant here as we are not estimating h2 from relatedness. Pop stratification is captured by the LDSC intercept. MAF and LD heterogeneity are important when thinking about which reference LD scores to use. We assume that the MAF and LD scores are similar (ideally identical) in the GWAS sample and the reference panel where the LD scores were calculated. That’s why the LD scores should come from the same ancestry group as the GWAS sample