30 F R E D M. U B E R 



numerous causes, both known and unknown, it is almost essential to 

 resort to a statistical analysis. That proper evaluation of the data is 

 not always made has been shown quite strikingly in a survey by 

 Dunn {12), who analyzed 200 articles in the field of medical physiol- 

 ogy, selected at random from American journals prior to 1930. 

 Dunn concluded: 



"(a) In over 90 per cent statistical methods were necessary and not used. 



"(b) In about 85 per cent considerable force could have been added to the 

 argument if the probable error concept had been employed in one form or 

 another. 



"(c) In almost 40 per cent conclusions were made which could not have 

 been proved without setting up some adequate statistical control. 



"(d) About half of the papers should never have been published as they 

 stood : either because the numbers of observations were insufficient to prove 

 the conclusions or because more statistical analysis was essential." 



This disgraceful indictment refers to a condition which existed twenty 

 years ago, but there is still reason to stress the value of a statistical 

 analysis of research, data even though scientific workers generally 

 have become increasingly statistics-conscious in the past two decades. 

 If such a waste of research time and money as indicated in the study 

 by Dr. Dunn were to continue, it might ultimately lead adminis- 

 trators to impose regulations that would not only curtail freedom of 

 investigation but also stifle research initiative. 



Statistical methods furnish a valuable tool in the conduct of ex- 

 perimental research for a number of reasons. CompHcated sets of 

 observations often can be described briefly and in simple terms. The 

 use of an objective criterion of accuracy furthers a critical attitude 

 toward experimental techniques and serves as a check on their re- 

 liability. Distinctions between significant and nonsignificant experi- 

 mental differences can be established in terms of an objective tech- 

 nical index. Thus meaningless and unwarranted conclusions can often 

 be avoided, saving both embarrassment and the time of other workers 

 in fruitless attempts to confirm the results. Not the least important 

 benefit from a careful statistical analysis of research data is the con- 

 fidence created in the results on the part of the original investigator. 



Statistical methods, as valuable as they are conceded to be, are 

 not a panacea for all the ills that befall experimentation. Their use 

 demands that the original data be of high quality. The observations 

 must be made meticulously and recorded accurately, with neither 

 sins of omission nor commission. It may be not only valueless but 



