Sequence data and Y chromosome or mtDNA haplotypes

The structure model assumes that loci are independent within populations (i.e., not in LD within populations). This assumption is likely to be violated for sequence data, or data from non-recombining regions. The impact of using such data is likely to be that (1) the algorithm underestimates the degree of uncertainty in ancestry estimates, and in the worst case, may be biased or inaccurate; (2) estimation of $K$ is unlikely to perform well. One valid solution is to recode the haplotypes from a linked region so that it is represented as a single locus with $n$ alleles. If there are very many haplotypes, one could group related haplotypes together. We are also aware of analyses that have taken the polymorphic sites from sequence data from multiple regions, and treated these within structure as independent loci. This type of analysis may yield sensible and informative results, however considerable caution must be applied to interpreting the results. The linkage model is likely to perform better here than the independent sites model. We would not recommend this type of approach for sequences from just one or a few regions, except perhaps in purely exploratory analysis.