Linkage Disequilibrium Influence on Regression Results


Suppose you have two SNP’s in linkage disequilibrium with a D’ value near 1 but a low r^2 value near 0.1. This essentially means that the alleles are almost always inherited together, but can’t be used to impute the other locus due to low alternate allele frequencies (i.e. they do not contain the same information most of the time).

Could it be argued that because the SNP’s do not contain identical information (and thus are not strongly collinear) that their regression coefficients could be reasonably assumed to be mostly independent of each other? Could it be argued that populations with higher r^2 values provide more biased estimates of effect at either loci than a population with a lower r^2 value (assuming D’ is constant and still near 1)?