THE ART OF SCIENTIFIC INVESTIGATION 



usually concerned with carefully pre-arranged experiments set up 

 to test an idea already bom. Also, in trying to provide sufficient 

 data for statistical analysis, the experimenter must not be tempted 

 to do so at the expense of accurate observation and of care w^ith 

 the details of the experiment. 



The use of statistics does not lessen the necessity for using 

 common sense in interpreting results, a point which is sometimes 

 forgotten. Fallacy is especially likely to arise in dealing with field 

 data in which there may be a significant difference between two 

 groups. This does not necessarily mean that the difference is 

 caused by the factor which is under consideration because 

 possibly there is some other variable whose influence or import- 

 ance has not been recognised. This is no mere academic possi- 

 bility, as is shown for example by the confusion that has arisen 

 in many experiments with vaccination against tuberculosis, the 

 common cold and bovine mastitis. Better hygienic measures 

 and other circumstances which may influence the results are 

 often coupled with vaccination. Statistics may show that people 

 who smoke do not on the average five as long as people who do 

 not smoke but that does not necessarily mean that smoking 

 shortens life. It may be that people who do not smoke take more 

 care of their health in other and more important ways. Such 

 fallacies do not arise in well designed experiments where the 

 initial process of randomisation ensures a valid comparison of 

 the groups. 



The statistician, especially if he is not also a biologist, may be 

 inclined to accept data given him for analysis as more reliable 

 than they really are, or as being estimated to a higher degree of 

 accuracy than was attempted. The experimenter should state 

 that measurements have been made only to the nearest centi- 

 metre, gram or whatever was the unit. It is helpful for the 

 statistician to have had some personal experience of biological 

 experimentation and he ought to be thoroughly familiar with all 

 aspects of experiments on which he is advising. Close co-opera- 

 tion between the statistician and the biologist can often enable 

 enlightened common sense to by-pass a lot of abstruse mathe- 

 matics. 



Occasionally scientific reports are marred by the authors 

 giving their results only as averages. Averages often convey 



22 



