246 Bulletin 397 



which the studies were made. In the first place, the continuous system of 

 conducting feeding experiments was the method under consideration. 

 Secondly, an experimental period of eighteen weeks was arbitrarily 

 chosen, for reasons already mentioned. Finally, use was made of the 

 records of the general college herd, not of animals on experiment. The 

 reasons for using such a group of animals have been given. 



It is believed that data are presented that will make possible a more 

 intelligent use of the continuous system of feeding experiments. These 

 data should aid the investigator in planning experiments so as to obtain 

 results capable of interpretation. The results given here should aid in 

 determining what variation may be expected under the conditions of 

 selection possible in a given case. It is further believed that these studies 

 should convince workers of the necessity of a statistical analysis of results 

 before conclusions are drawn. 



Variation of similar individuals and groups, even when selected care- 

 fully and treated uniformly, is a fact to be acknowledged and considered. 

 Hence, in an experiment in which two or more groups of animals of similar 

 productive capacity are used, one group will always produce more than 

 the others. This may be due to the causes considered in the experiment, 

 or it may be in spite of them. In such experiments, conclusions should not 

 be drawn from inconclusive results. 



Statistical analysis of the results of feeding experiments is an aid in 

 interpreting the results. The greatest advantage of the statistical method 

 is that it eliminates personal bias and preconceived conclusions, and 

 gives the mathematical chances for or against a proposition. Even odds 

 of 30 to I in favor of a conclusion do not give absolute certainty. The 

 statistical method of analysis will not take the place of judgment, but it 

 is a valuable adjunct to judgment in evaluating results and formulating 

 conclusions. 



It cannot be determined how closely the figures found for individual 

 variation over a period of eighteen weeks would apply to periods of other 

 durations. It is believed that for periods of approximately the same 

 length the application would be very close, while for periods of much 

 longer duration the value reported should at least serve as a guide. In 

 considering the application of the results as to the influence of certain 

 selective factors on variation, their value is not limited by the fact that a 

 period of definite length was used. Rather, these results should have a 

 general application. 



In making up groups of animals for feeding experiments the following 

 points should be considered: 



I. Groups should be selected on the basis of production for a trial 

 period just preceding the experimental period. 



356 



