Hi - We are interested in using the by-chromosome analysis to estimate undetected population substructure as described in several GCTA papers -- testing heritability for each chromosome separately, and then within a single analysis, and regress the chromosome-specific heritabilities from the separate analyses onto the combined analyses. But we don't quite see how to implement the latter (autosome-wide analysis, but estimating heritability for each chromosome within that analysis). Could you advise? Thanks.
1) Calculating 22 GRMs with each based on all SNPs on one chromosome (--make-grm) 2) Running a REML analysis fitting the each of GRMs one by one (--reml and --grm) 3) Running the REML analysis fitting all the GRMs simultaneously (--reml and --mgrm)