EXPERIMENTATION 



When testing out a procedure for the first time it is often 

 impossible to estimate in advance how many animals are required 

 to ensure a decisive result. If expensive animals are involved 

 economy may be effected by doing a test first with a few animals 

 and repeating the test until the accumulated results are sufficient 

 to satisfy statistical requirements. 



One of the basic conceptions in statistics is that the individuals 

 in the group under scrutiny are a sample of an infinitely large, 

 hypothetical population. Special techniques are available for 

 random samphng and for estimating the necessary size of the 

 sample for it to be representative of the whole. The number 

 required in the sample depends on the variability of the material 

 and on the degree of error that will be tolerated in the results, 

 that is to say, on the order of accuracy required. 



Fisher considers that in the past there has been too much 

 emphasis placed on the importance of varying only one factor 

 at a time in experimentation and shows that there are distinct 

 advantages in planning experiments to test a number of variables 

 at the same time. Appropriate mathematical techniques enable 

 several variables to be included in the one experiment, and this 

 not only saves time and effort, but also gives more information 

 than if each variable were treated separately. More information 

 is obtained because each factor is examined in the light of a 

 variety of circumstances, and any interaction between the factors 

 may be detected. The traditional method of experimental isola- 

 tion of a single factor often involves a somewhat arbitrary 

 definition of that factor and the testing of it under restricted, 

 unduly simphfied circumstances. Complex, multiple factor experi- 

 ments, however, are not so often applicable to work with animals 

 as to work with plants, although they can be used with advantage 

 in feeding trials where various combinations of several com- 

 ponents in the ration are to be tested. 



Statistics, of course, like any other research technique, has 

 its uses and its limitations and it is necessary to understand its 

 proper place and function in research. It is mainly valuable in 

 testing an hypothesis, not in initiating a discovery. Discoveries 

 may originate from taking into consideration the merest hints, 

 the slightest diflferences in the figures between different groups, 

 suggesting something to be followed up; whereas statistics are 



21 



