Constrained Least Squares Estimation of Mixed Population Stock Composition from mtDNA Haplotype Frequency Data

Nuclear allozymes have provided the markers of choice for genetic stock identification, especially for Pacific salmonids. For other species, a dearth of allozyme variation has led workers to pursue mtDNA markers. Haploid mtDNA markers obviate methods based on Hardy–Weinberg equilibrium; all mtDNA markers are rigidly linked, violating the independence assumption; and mtDNA sample sizes are small. Based on a conditional weighted least squares approach for the case of known source population genotype frequencies, we develop an unconditional method for estimated genotype frequencies. We compare the methods with Monte Carlo simulation, evaluating effects of source population and mixed harvest sample sizes and those of source population divergence and mixture composition. When genetic resolution is poor or sample sizes are inadequate, the conditional method works better, but with larger sample sizes and reasonable population resolution, the unconditional method is superior.