Pella and Masuda: Bayesian methods for analysis of stock mixtures from genetic markers 
159 
0.3 
0.3 
0.2 
0.2 
0.1 
0.0 
.■III.. 
0.1 
0.0 
6 
I 
0 10 20 30 40 50 0 10 20 30 40 50 
0.3 
0.2 
0.1 
0.0 ' 
0 10 
20 
30 
40 50 
0.3 
0.2 
0.1 
0.0 
20 
30 
40 50 
30 40 50 
Type count 
Figure 1 
Histograms of predictive baseline population (Gulf of Maine-Bay of Fundy, top; Gulf of St. Lawrence, second; New- 
foundland, third; and West Greenland, bottom) sample counts for the most common haplotype in the pooled summer 
and winter samples, by empirical Bayes (left) and pseudo-Bayes (right) methods. The actual count of the haplotype 
in each baseline sample is shown as a spike. 
Table 1 
Parameters of the posterior density for harbor porpoise population proportions composing the winter stock mixture. Reported pro- 
portions do not necessarily sum to 1.0 because they are rounded. 
Population 
Mean 
Mode 7 
SD 
2.5% 
Posterior quantiles 
Median 
97.5% 
Gulf of Maine-Bay of Fundy 
0.12 
0.02 
0.13 
0.00 
0.08 
0.46 
Gulf of St. Lawrence 
0.48 
0.69 
0.19 
0.14 
0.48 
0.84 
Newfoundland 
0.15 
0.02 
0.16 
0.00 
0.10 
0.52 
West Greenland 
0.24 
0.26 
0.18 
0.00 
0.22 
0.66 
7 The mode is computed by 4-dimensional binning of the Markov chain Monte Carlo samples for stock proportions, each bin with sides of 0.05, and 
then normalizing the bin center having maximum count. 
percentile interval (Efron and Tibshirani, 1993) for confi- 
dence bounds. The alternate method, called the non- 
symmetric percentile bootstrap (Lunneborg, 2000), is 
expected to have superior coverage properties to the stan- 
dard percentile method for the usual skew distributions 
of stock-mixture composition estimates. The confidence 
