228 



Bulletin 397 



therefore about 54 to i that a repetition of the study would give similar 

 results. Wood and Stratton (19 10) have suggested that for data obtained 

 from agricultural experiments, odds of 30 to i may be considered as 

 conclusive evidence. Odds of 30 to i require that the difference must be 

 3.17 times its probable error. The data given in this case more than 

 satisfy the requirements suggested by Wood and Stratton, and it may 

 therefore be concluded that groups of cows of similar individual produc- 

 tion are less variable than groups of unselected animals. 



As a further test of the effect of selecting equal individuals, groups 

 were made up whose total production for two weeks was from 3200 to 

 3299 pounds but which were divided into three classes according to indi- 

 vidual production. The individuals of Class I had productions of from 

 500 to 599 pounds for the trial period; those of Class II had productions 

 of from 400 to 499 pounds or from 600 to 699 pounds; those of Class III 

 had productions under 400 pounds or over 699 pounds. By means of 

 this method of selection, there were made available groups of six cows 

 of similar total production, divided into three classes according to the 

 similarity of the individuals as shown by their production during the 

 trial period. The results of this selection are shown in table 4 : 



TABLE 4. Effect of Similarity of Individual Production on Variability of 



Groups 



Class 



I 



II 

 III 



Individual production, 



two- weeks period 



(pounds) 



500-599 



400-499, 600-699 

 Below 400, above 699 



Group production, 



two-weeks period 



(pounds) 



3,200-3,299 

 3,200-3,299 

 3,200-3,299 



Number 

 of groups 



9 

 II 

 12 



Coefficient of 

 variability 



2 . 98 ± o . 47 

 4 . 84 ± o . 70 

 5.27 ± 0.73 



The difference in variability between the groups of Class I and those 

 of Class II is 1.86 ± 0.84. Here the difference is 2.21 times its probable 

 error, and the odds are 6 to i that the difference in variability is due to 

 the factor of selection. The difference in variability between the groups 

 of Class I and those of Class III is 2.29 ± 0.87. Here the difference is 

 2.63 times its probable error, and the chances are 12 to i that groups of 

 individuals of similar productive capacity are less variable. The con- 

 sistency of the results listed in tables 3 and 4 seems to indicate that sim- 

 ilarity of individuals in productive capacity has an important effect on 

 variability of the group, and that in grouping cows for a feeding test the 

 experimental error can be reduced by using similar individuals. 



338 



