FISHERY BULLETIN: VOL. 86, NO. 3 



Table 3.— Statistical significance of annual changes in 

 relative population number of sablefish, Gulf of Alaska, 

 Japan-U.S. cooperative longline survey, 1979-86. The sym- 

 bols used are defined as follows; + signifies a significant 

 increase in RPN; - signifies a significant decrease in 

 RPN; O signifies no significant change. 



mates of catch per hachi as if the strata had been 

 sampled. Because the estimated values mimic the 

 observed data, the probability of rejecting the null 

 hypothesis is somewhat higher than the nominal 

 value. In general an inaccurate estimate should have 

 a small total effect because catch per hachi values 

 were missing for only 9 of the 320 strata available 

 during the survey (8 years x 5 areas x 9 strata). 

 The only comparison where the estimated catch per 

 hachi values might have caused an incorrect conclu- 

 sion is the 1984-85 comparison where 7 strata were 

 unsampled. This comparison was retested using only 



the observed catch per hachi values. The annual 

 change, like the comparison using the estimated 

 catch per hachi values, was found to be statistically 

 significant. 



Another potential effect of incomplete sampling 

 at a station is bias in the bootstrap calculation. Each 

 bootstrap replicate RPN^j^* was computed from a 

 bootstrap sample selected by sampling with replace- 

 ment from the stations within an area. If one or 

 more stations were missing an observation for a 

 strata, then it was possible for the bootstrap sam- 

 ple to have all missing values for the strata. In this 

 case, no catch per hachi values were available to 

 calculate a value for the strata and the value for the 

 strata was treated as zero. This treatment may 

 negatively bias the resultant RPN, ;,*. The extent 

 of the negative bias was tested by comparing the 

 RPN to a bootstrap estimate of the RPN, denoted 

 RPNfc, where RPN^ is the mean of the 1,000 

 RPNijt* from an area and year. If RPN^ generally 

 was smaller than RPN, then the bootstrap replicates 

 were negatively biased. The percentage difference 

 between the resultant RPNj's compared to the 

 RPN's ranged from - 0.7 to + 0.5 and averaged only 

 -1-0.03, showing that the negative bias had little 

 effect on the RPN,^*. 



Other fishery and survey data substantiate the 

 significance of the overall increase in the RPN and 

 also the marked increases from 1981 to 1982 and 

 from 1984 to 1985. The CPUE in the Japanese long- 

 line fishery showed a similar pattern to the RPN's 

 for the years the data overlapped, from 1979 to 

 1983; the fishery CPUE increased from the late 

 1970's, with the largest increase from 1981 to 1982, 

 and decreased from 1982 to 1983 (Fig. 4). 



Examination of length compositions for depths 

 101-200 m indicate that the strong 1977 year class 

 (Sasaki 1982; McFarlane and Beamish 1983; Funk 

 and Bracken 1984) was responsible for the RPN in- 

 crease from 1980 to 1982 and that the RPN increase 

 from 1984 to 1985 was due to a strong 1980 year 

 class. Sablefish recruiting to the survey area first 

 appear at depths 101-200 m, and strong year classes 

 are more distinguishable in the length compositions 

 at these depths. The length compositions for depths 

 101-200 m show the initial appearance and subse- 

 quent increase in length of fish of the strong 1977 

 year class (Fig. 5). The mode at 47 cm FL in 1979 

 indicates the first year that the 1977 year class was 

 available to the survey gear. The rightward move- 

 ment of the mode in succeeding years illustrates the 

 increase in fish length for the 1977 year class. The 

 movement of the mode slowed in 1982 which we in- 

 terpret to be due mainly to the movement of larger 



450 



