Closed jdblischak closed 3 years ago
In practice, LDSC will also preprocess the sumstats files to give reliable heritability and genetic covariance estimates. Therefore, LDSC will give similar results either the original sumstats files or the preprocessed sumstats files are passed to argument --sumstats
. Yet I agree that it would be more straightforward passing the preprocessed sumstats files to LDSC.
In Step 1, the two original sumstats files are preprocessed to create
output_dir/Data_QC/dat1.txt
andoutput_dir/Data_QC/dat2.txt
(and the per chromosome files). Then in Step 2, the original sumstats files are passed to the LDSC argument--sumstats
. Shouldn't the preprocessed sumstats files be passed to LDSC? I realize that in the tutorial example these files are identical (i.e.preprocess.R
has no effect), but if it did filter some SNPs or flip some alleles and effect sizes, shouldn't they be used?