a much smaller standard error, and since the 

 underlying data usedare essentially similar, it 

 seems inappropriate to use relation (2) and the 

 prediction derived from it. 



Regression of Returns at Ages 3 and 4 on 

 Mean Air Temperature 



The data for a regression of returns at 

 ages 3 and 4 on mean tennperature (table 23) 

 include as a predictive variable the annual 

 mean air temperature at St. Paul Island for 

 the 12-month period ending 30 June of the 

 year of birth of the year class in question. 



The resulting regression equation is: 



K = 15,800 + 10, SOOT (r = 0.80) 



^ For the 1963 year class T = 2.8 and hence 

 K = 45,500. 



The kill of 3-year-olds in 1966 was 25,500, 

 so this estimate implies a balance of 20,000 

 4-year-olds. The standard error of the pre- 

 dicted K is 9,700, which was also the standard 

 error of the forecast of the kill of 4-year-old 

 males in 1966 by this method. 



Regression of Returns at Ages 3 and 4 on 

 Estimated Number of Yearlings 



Since seals of the 1963 year class were not 

 tagged as yearlings in 1964, a forecast based 

 on tag recoveries from yearlings cannot be 

 made. 



Table 23. --The kill ot 3- and 4-year-old male seals and meaii 

 air temperature, year classes 1950-62, St. Paul Island 



Combined Estimates 



The two most satisfactory estimates and 

 their S.E. (standard errors) are: 



1. From regression on kill of 3-year-olds 

 and the mean round: 



Estimate 14,000 



S.E. 1,800 



Since the terminal date of killing in 1966 was 

 5 August, no adjustment is required in the K 

 obtained from the regression equation. 



2. From regression on mean air tempera- 

 ture: 



Estimate 20,000 



S.E. 9,700 



The combined estimate is 14,300; since the 

 weighting was inversely proportional to the 

 square of the standard error, almost all the 

 weight was attached to the apparently very 

 reliable first estimate. 



FORECAST OF KILL OF 3 -YEAR -OLD 

 MALES ON ST. PAUL ISLAND 



This section describes three methods cur- 

 rently used to forecast the kill of 3-year-old 

 males. They are: (1) Regression of mean 

 air temperature on return; (2) regression of 

 the kill of 3-year-old male seals on weight 

 of pups; and (3) regression of the percentage 

 of year class taken in the harvest onthe dead- 

 pup count percentage. Estimates of the number 

 of yearling males and their potential use in 

 predicting the size of the kill at age 3 are 

 also discussed. 



Regression of Mean Air Temperature 

 on Return 



As shown in an earlier section, the regres- 

 sion of the kill (K) at ages 3 and 4 on the 

 annual mean temperature (T) of the year 

 class for the 12-month period prior to 30 June 

 is: 



k = 15.800 + 10,600T 



This equation is based on data for 13 year 

 classes. For the 1964 year class that will 

 contribute 3-year-olds in 1967, T was 1.5 

 (the mean temperature was 33.45°) which 

 implies that K = 31,700. The standard error 

 is 9,700. If 67 percent of this estimated total 

 is taken at age 3, then the forecasted kill at 

 age 3 in 1967 is 21,200. 



Regression of the Kill of 3 -year-old 

 Male Seals on Weight of Pups 



The data for the regression of the kill at 

 age 3 on St. Paul Island on the weight of 

 untagged pups in autumn are shown in table 24. 



27 



