Refinement of Feeding Experiments for Milk Production 235 



The coefficients of variability in table 1 1 show the same general range 

 as those in table 10. The average value is also similar. If there is any 

 difference due to the different systems of determining the limits, the 

 coefficients should be larger where the range is greater in proportion to 

 the group yield. In table 10 the limits are wider below 500 pounds, 

 while above this figure the converse is true. Splitting each table at this 

 point, the average coefficient for the groups below 500 pounds is 

 12.89 =t 0.58 for table 10 and 11. 91 ± 0.54 for table 1 1 ; above 500 pounds 

 the average coefficient is 11.08 ± 0.58 for table 10 and 11.49 =t 0.65 

 for table 1 1 . Thus it is seen that the coefficients are slightly larger where 

 the limits are wider, but the probable errors indicate that the differences 

 are not significant. It therefore appears that there is no object in using 

 a percentage basis of limits under the conditions of selection used. 



Allowable range in selecting individuals of equal production. — In the 

 studies previously made in selecting individuals of equal production, 

 the range allowed has been 50 pounds. A study was made to determine 

 the effect of a wider and of a narrower range on variation, in order to 

 ascertain whether other limits would prove more satisfactory. First, 



TABLE 12. Range to be Allowed in Selecting Individuals of Equal Production 



Individual production, two-weeks period 

 (pounds) 



400-499 



400-449 



450-499 



400-424 .... 



425-449 



450-474 . . . . 

 475-499 • • • • 



400-412.4. . 

 412.5-424.9 

 425-437-4 ■ • 

 437 -5-449 -9 

 475-487.4. . 



Range 



allowed 



by limits 



(pounds) 



100 



50 



25 



12.5 



Number of 

 individual? 



51 



51 



51 



34 



Coefficient of 

 variability 



12.65 ±0.84 



1 2 . 03 ±0.81 



10.98 dz 0.74 



10.68 ±0.89 



all the cows producing from 400 to 499 pounds in the trial period were 

 selected and their coefficient of variability was determined. Thus this 

 measure of variation was based on the assumption that a range of not 

 over 100 pounds in two weeks represented equal production. Next, 

 the same animals were divided into two classes, one composed of indi- 



345 



