BRISTOL BAY AND ALASKA PENINSULA SALMON STATISTICS 



63 



statisticians generally agree that a 

 coefRcient that is three times its 

 probable error is significant of some 

 degree of association. It is possi- 

 ble, of course, that some factors 

 other than a predominance of five- 

 year fish has caused this high cor- 

 relation between catches at five- 

 year intervals, but we have no sug- 

 gestion to make as to what these 

 factors may be. 



In the Nushagak district there 

 is an imdoubtedly significant corre- 

 lation between the size of the 

 catches at four-year intervals; the 

 coefficient is over six times its 

 probable error. It is apparent from 

 the graph, however, and also from 

 the work sheets made in the process 

 of calculating the correlation coeffi- 

 cients, that the correlation between 

 the catches at four-year intervals 

 is due mainly to an exceptionally 

 close association, which has been 

 maintained in comparatively recent 

 years since about 1914. Previous 

 to this time there was much more 

 of a tendency toward correlation in 

 catches at five-year intervals, but 

 the strong tendency toward correla- 

 tion at four years, which has pre- 

 vailed recently, has, in considering 

 the entire series, entirely out- 

 weighed the earUer condition. The 

 correlation between catches at five- 

 year intervals in the Nushagak dis- 

 trict is not significant, the coeffi- 

 cient being less than its probable 

 error and, as would necessarily be 

 the case with a strong correlation 

 at the four-year interval, the cor- 

 relation at the six-year interval is 

 significantly negative. Such a neg- 

 ative coefficient of correlation means 

 that in general a good catch in a 

 100621—28 4 



Fio. 8.— Percentage fluctuations trom the trends of the catches of red 

 salmon in the four districts of Bristol Bay 



