22 2 Bulletin 397 



remain a certain number of uneliminated factors, the probable limits of 

 variation of which must be known for the intelligent interpretation of 

 results. Thus statistical data relative to the variation in milk production 

 under various conditions of selection are needed. 



STATISTICAL STUDY OF V.^RIATION IN MILK PRODUCTION 



OBJECT AND METHODS 



The object of this statistical study, as has already been pointed out, 

 was to obtain the data needed for the intelligent planning of feeding 

 experiments conducted under the continuous system. These data con- 

 sist primarily of information as to the probable limits of variation, over 

 a given experimental period, of individuals and groups selected on the 

 basis of equal productive capacity as determined in a preliminary period. 

 It has been pointed out that variation can be lessened by so selecting 

 the animals as to eliminate certain controllable factors affecting milk 

 production, and that the niunber of factors which can thus be fixed is 

 determined by the number of available animals from which the selection 

 is made. Therefore data as to the effect of various selective factors on 

 variation are needed, to attain the object of this statistical study. 



The data for this study were obtained from the records of the university 

 herd. These records are complete over a period of thirty years, and com- 

 prise data for some four hundred individuals. For the purpose of this 

 study it would be desirable in some respects to have all the records from 

 one calendar year, in order that all animals might be under the most 

 uniform conditions of feeding and environment. This was of course impos- 

 sible because of the large niunber of animals required for the study. How- 

 ever, the care and management of the university herd have been in accord- 

 ance with the same general principles for the period of years represented 

 by the data used. The records have made possible the exclusion from 

 the study of all animals that have been used in any feeding experiment. 

 It may be stated that animals on experiment are accorded more uniform 

 treatment than the general herd, which is used for teaching purposes, 

 and a greater degree of exactness woiild result if the data were obtained 

 from records of such animals. But it was obviously impossible to make 

 the present study from experimental records. In fact, if the data were 

 thus obtained, they would be applicable only to investigations in which 

 an experimental herd, separate from the general herd, was available. 

 It is therefore believed that the data obtained in the present study have 

 a wider application. 



The breeds used in making this study were the Holstein-Friesian and 

 the Jersey. The greater proportion of the work was done with the former 



332, 



