850 
Fishery Bulletin 96(4), 1998 
inheritance at all three loci, although a null (nonamp- 
lifying) allele is present at O/sl03 (Small et ah, 1998). 
Individual samples that although amplified with both 
other primer sets yet produced no Ots 103 alleles af- 
ter the Ots 103 amplification was performed three 
times, were scored as Ots 103 null homozygotes. Al- 
lele frequencies at the Ots 103 locus were then cor- 
rected according to a maximum-likelihood estimate 
(EM algorithm, Dempster et ah, 1977) by using 
GENEPOP, version 1.2 (Raymond and Rousset, 
1995a). Corrected frequencies were used in the ge- 
netic distance analysis. For OtslOl and OtsS, fish with 
a single allele were scored as homozygotes and fish with 
two alleles were scored as heterozygotes. Allele fre- 
quency data for individual populations in the Fraser 
River and Thompson River are from Small et al. ( 1998) 
and only the weighted means of total allele frequen- 
cies for these regions are presented here ( Tables 1 and 
2). The upper Fraser River population (Bridge River) 
has been grouped with the Thompson River popula- 
tions in accordance with its genetic profile. 
Hardy-Weinberg equilibrium (HWE) was tested 
with the probability test of HWE (an exact HW test, 
Guo and Thompson, 1992) with GENEPOP (version 
1.2). F-statistics and their standard deviations were 
calculated according to Weir and Cockerham (1984) 
using FSTAT (Goudet, 1995). We use the notations 
F st , F s and F j( for Weir and Cockerham’s 9, f, and F, 
respectively. P-values for a = 0.05 in all data analy- 
ses were corrected for simultaneous multiple tests 
(Lessios, 1992). Pairwise comparisons of populations 
for differences in allele distributions were conducted 
at all three loci with Fisher’s exact tests (Raymond 
and Rousset, 1995b). In populations with multiple 
year classes, annual variability in allele frequencies 
was tested with 1000 simulations in a Monte Carlo 
analysis (Roff and Bentzen, 1989) and populations 
were divided into year classes for a neighbor-joining 
(NJ) analysis (Saitou and Nei, 1987). A NJ dendro- 
gram illustrating genetic relationships among popu- 
lations was constructed with PHYLIP 3.5c software 
(Felsenstein, 1993). The allele frequency matrix was 
resampled 500 times and Cavalli-Sforza and 
Edward’s (1967) chord distances were estimated for 
population pairs. NJ dendrograms were constructed 
for each matrix and a consensus NJ dendrogram was 
generated with CONSENSE (see Felsenstein, 1993). 
Individual fish were classified to specific populations 
with a jackknife discriminant analysis (SAS Insti- 
tute Inc., 1989). 
Estimation of stock composition 
Microsatellite allele frequency data were examined 
for use in estimating stock composition in a mixed- 
stock fishery. We pooled low-frequency alleles in ad- 
jacent bins (so that each bin contained at least 6% of 
the alleles) to reduce the number of genotypes for 
which frequencies were estimated. This pooling re- 
sulted in 13 “analysis” bins (91 genotypes) for Ots 101, 
10 bins (65 genotypes) for Ots 3, and 13 bins (91 geno- 
types) for Ofsl03 (Table 3). Genotype frequencies at 
Ots 101 and Ots 3 were estimated for each population 
from the allele frequencies by assuming a Hardy- 
Weinberg distribution of genotypes. Because geno- 
type frequencies at the Ots 103 locus were generally 
not in HWE owing to the presence of the null allele, 
the observed genotype frequencies were used to char- 
acterize each population for Ots 103. Genotype fre- 
quencies at all three loci were used as input in a 
maximum likelihood mixed-stock fishery analysis 
(Fournier et al., 1984). 
Two types of hypothetical mixed-stock fishery 
samples were simulated: single-region fishery 
samples composed of fish from lower Fraser River 
and the Thompson River coho salmon populations, 
and multiregion fishery samples composed of fish 
from populations from several regions. Populations 
contributing to the multiregion samples were cho- 
sen on the basis of known or inferred migration pat- 
terns. For the Fraser River fishery simulation, only 
populations from the lower Fraser and Thompson 
rivers were present in the baseline and in the mixed- 
stock fishery samples. In the multiregion fishery 
simulations, all 34 populations were used in the 
baseline. In all simulations, fishery samples of 200 
fish were generated from specified populations in the 
Table 3 
Method of pooling low-frequency alleles for OtslOl, Ots 3, 
and Otsl03 to reduce the number of genotypes for baseline 
populations in mixed-stock analyses. 
Microsatellite allele bin numbers 
Analysis bin no. 
OfslOl 
Ots 3 
Ots 103 
1 
1-7 
1 
1, 2, 3 
2 
8, 9 
3-11 
4-7 
3 
10 
12 
8-11 
4 
11 
13 
12 
5 
12 
14 
13 
6 
13, 14 
15 
14, 15 
7 
15, 16, 17 
16 
16-20 
8 
18, 19 
17 
21, 22 
9 
20, 21 
18 
23, 24 
10 
22 
19 
25, 26, 27 
11 
23, 24 
28,29 
12 
25, 26 
30-38 
13 
27-32 
39 
