554 



Fishery Bulletin 104(4) 



given in Table 1 (reproduced from Koljonen et al., 2002). 

 An estimate of total F^j, (Weir and Cockerham, 1984) 

 equal to 0.07 was obtained in Koljonen et al. (2002) 

 for these stocks on the basis of the nine microsatellite 

 loci. The magnitude of the genetic differentiation in the 

 underlying population is fairly small, and the pairwise 

 distances vary considerably. Thus, we may conclude that 

 these stocks represent a biologically challenging setting 

 for inference about the genetic mixture in a population 

 sample. Using the individual stock allele frequencies, 

 we have simulated baseline individuals and catch sam- 

 ples under the assumptions of HWE and no linkage 

 between the loci. A wide variety of configurations, with 

 complete and partial baseline information and different 

 sample sizes, were tested. In the analyses involving five 

 underlying stocks, we used Ti' = 10 as the prior upper 

 bound, and the estimation algorithm was run 12 times 

 for each replicate data set. For cases with only two un- 

 derlying stocks, the upper bound was set as K = 6. 



Results of our simulation experiments are summa- 

 rized in Tables 2-8. As a summary, we highlight the 

 following aspects. Uneven proportions of stock presence 

 in the samples do not seem to affect the inference nota- 

 bly, even when the baseline information is only partial. 

 The results in Table 3 are produced under a particu- 

 larly challenging situation, where the baseline informa- 

 tion comprises 40 individuals from a single stock only. 

 The sample configuration then contains 40 individuals 

 from this stock and 10 individuals from another, a pri- 

 ori unknown stock. The results show that our method 

 performs surprisingly well in the identification of the 

 outgroup, given that the genetic difference between 



the two underlying stocks is not negligible. However, 

 as the results in Table 4 illustrate, the presence of pu- 

 tative stocks not represented by baseline information 

 may also be masked by the baseline available for a ge- 

 netically similar stock. Identification of putative stocks 

 without using any baseline information may de facto 

 be more successful under such circumstances (compare 

 Tables 4a-4d). Therefore, we suggest that in practice 

 both types of analyses are performed and the results 

 compared, since this is computationally inexpensive 

 with our method. Our results indicate that incomplete 

 baseline information is expected to be most fruitful for 

 the identification task when there are baseline samples 

 available from the stocks that are genetically most 

 similar. The baseline configurations in Table 4 can be 



