REISENBICHLER and PHELPS: GENETIC VARIATION IN CHINOOK AND COHO SALMON 



Data Analysis 

 Goodness-of-Fit Tests 



We used the chi-square test to examine geno- 

 type frequencies for deviation from the (Hardy- 

 Weinberg) proportions expected with random 

 mating. Cells with an expected number <5 were 

 combined with the next larger cell. The signifi- 

 cance level for each test was modified to account 

 for the increase in type I error when multiple 

 tests of the same hypothesis are made (Cooper 

 1968). Tests were considered significant if the chi- 

 square statistic exceeded the critical value for 

 chi-square associated with a probability of 0.05/n , 

 where n was the number of loci tested within a 

 sample. In this way the overall probability of re- 

 jecting Hq by chance alone was approximately 

 1 - (1 - 0.05/;?)" = 0.05 for each sample. Geno- 

 types for Idh-3,4, Mdh-1^, and Mdh-3,4 were not 

 tested because these systems consisted of pairs of 

 loci with identical electrophoretic mobility, and 

 genotypes at each locus could not be determined. 



The likelihood ratio test (G-test; Sokal and 

 Rohlf 1981) was used to test equality in allele 

 frequencies between year classes. Here also, cells 

 with an expected number <5 were combined with 

 the next largest cell. The G -statistics, summed 

 over all loci, were considered significant if they 

 exceeded the critical value for chi-square associ- 

 ated with a probability of 0.05/s, where s was the 

 number of samples tested. Samples from streams 

 and samples from hatcheries were tested as sepa- 

 rate groups. The correction for multiple compari- 

 sons was made because each of the three Hq — no 

 interbrood variation by drainage, by streams 

 within drainages, or by hatchery — was independ- 

 ently tested for several drainages, streams, or 

 hatcheries, respectively. 



Analysis of Variance 



We used analysis of variance (ANOVA) to test 

 interdrainage differences, differences between 

 hatchery and wild chinook salmon, and differ- 

 ences between summer and fall runs of chinook 

 salmon. Data for coho salmon were not tested by 

 ANOVA because data were available for only one 

 year class from most locations, and estimates of 

 interbrood variation in allele frequencies would 

 have come from only two sample locations. The 

 data used were from the loci scored for fish from 

 each major north coast drainage and with fre- 

 quencies <0.95 for the common (100) allele. The 

 values used in the analysis were the arcsin of the 



square root of the frequency of the common allele 

 at each locus. Differences were tested by contrasts 

 (Table 3) or by partitioning the sum of squares 

 within a one-way ANOVA for each locus 

 (Snedecor and Cochran 1967; SPSS, Inc. 1983). 

 Groups included in this analysis were as follows 

 (adults would have spawned in 1983): 



Cell 



Group 



Run 



Replicate 



'Milner, G. B., D J. Teel, and F M Utter. 1983. Genetic stock iden- 

 tification study; final report of research Unpubl. Rep. Natl. Mar. Fish. 

 Serv., NOAA, Seattle, WA. 



Juveniles from the different runs of chinook 

 salmon were morphologically indistinguishable 

 and our estimates of error variance were probably 

 inflated because they were based on samples (of 

 juveniles) that vary from year to year in the pro- 

 portion of fish from each race. As a result, the 

 (discriminatory) power for detecting differences 

 between groups was impaired. In view of this re- 

 duced discriminatory power, differences with 0.05 

 < P < 0.1 were noted in the text; statistical sig- 

 nificance, however, was reserved for differences 

 with P < 0.05. 



Adult fall and summer chinook salmon from 

 the Quillayute River and adult fall chinook 

 salmon from the Quinault River were not in- 

 cluded in the ANOVA because adults returning 

 to these streams include large numbers of hatch- 

 ery fish (Houston fn. 3). Adult summer chinook 



685 



