Appendix table 3. 



--Analysis of variance of regression of adjusted kill on dead-pup count and 

 mean temperature, St. Paul Island 



~r 



Source of 

 variation 



Degrees 



of 

 freedom 



Sum of 

 squares 



Mean 

 square 



Dead -pup count and temperature 



Temperature alone 



Difference due to neglecting dead-pup count. 



Dead -pup count alone 



Difference due to neglecting temperature.... 

 Error ( residual ) 



2 



1 

 1 

 1 

 1 

 6 



1157.44 



1156.30 



1.14 



900.43 



257.01 

 414.12 



578.72 



1156.30 



1.14 



900.43 



257.01 



69.02 



.02 



3.72 



Three-year-old Males 



Method l .--The temperature-kill relation- 

 ship given above can be applied. The mean 

 temperature from July 1960 to June 1961 was 

 33.8, therefore, X = 18 and Y = 35.6. During 

 the past 5 years, 73 percent of the number 

 killed at ages 3 and 4 have been taken at age 

 3. On this basis, the predicted kill of 3-year- 

 olds is: 



(.73) (35,600) = 26,000 



Method 2 . --An alternative method of predic- 

 tion is based on average returns from the 

 1953-59 year classes against the average 

 number of females. By the accumulative esti- 

 mate 1 the average number of females age 3 

 and older was 801,100. The average kill of 

 3-year-old males from these year classes 

 was 28,500. This represents 3.56 percent of 

 the number of females. The accumulative 

 estimate of the number of females in 1961 

 is 610,000. Thus, anaveragekill would produce 

 21,700 or 22,000 animals. 



Confidence Intervals, Other Methods 



Confidence interval estimates for some of 

 the predictions can be obtained. For example, 

 the 95 percent confidence intervals for the 

 estimated adjusted kill from the 1960 and 

 1961 year classes, based on the temperature- 

 kill relationship are: 



I960: Y lies between 34,600 + 22,000, i.e., 

 12,600 to 56,600 



1961: Y lies between 35,600 + 22,000, i.e., 

 13,600 to 57,600 



These imply a confidence interval for the kill 

 of 4-year-old males in 1963 of from to 

 42,800. At present, the variability inherent 

 in these predictions is high. 



In other years, attempts were made to fore- 

 cast the kill of 3-year-old males on some 

 index from the kill of 2-year-old males the 

 previous season. Clearly, this is the best 

 approach. At present, however, the small 

 number of 2-year-old males killed and the 

 selection involved in their killing are factors 

 which prevent a satisfactory prediction. 



Prediction of Total Kill 



Three widely separated predictions for the 

 kill of 4-year-old males are available. Be- 

 cause information suggesting preference of 

 one prediction over another is not available, 

 the median value is used. Assuming that the 

 male kill will be terminated 5 August, the 

 resulting predictions for St. Paul Island in 

 1964 are: 



■'-An estimate based on the kill of males and on a sur- 

 vival rate for females 1.25 times that of males. 



An implied total male kill of 50,000 is based 

 on the usual kill ratio of 80:20 betweenSt. Paul 

 and St. George Islands, though the ratio in 

 1963 was 76:24 for 3- and 4-year-old males. 



39 



