St^^i^tic^ 



— a tool for the research man 



By SARGENT RUSSELL, Department of Agricultural Economics 



THE Startling conclusion that 

 men have more children than 

 women is an example of the gross 

 misuse of statistics. Yet this obvi- 

 ous inaccuracy grew out of the fact 

 that Yale men were recorded as 

 having more children than Smith 

 women. 



Although statistics is now an ac- 

 cepted research tool in all fields of 

 science, its proper use is still the big 

 question. 



Statistics Is a Tool 



Like any tool, statistics should 

 be used intelligently, and like some 

 beverages, with moderation. The 

 intelligent user of statistics knows 

 the capacity and limitations of his 

 tool. Too often a research worker, 

 looking for the royal road to statis- 

 tics, mechanically grinds out some 

 results. Then, without under- 

 standing his method or the signifi- 

 cance of the results, publishes them 

 with no interpretation or, worse 

 still, an incorrect one. 



One statistical procedure, corre- 

 lation, has residted in some classic 

 bungles because of improper inter- 

 pretation. Jevon's theory of busi- 

 ness cycles based on sunspots is out- 

 standing. Another worker made a 

 similar discovery by relating the 

 height of water in the Great Lakes 

 to the rate of business activity. 



When seeking a method of anal- 

 ysis for gathered data, intelligent 

 research men can prevent another 

 frequent blunder by looking to sta- 

 tistics before collecting any of their 

 data. 



On the other side of the pendu- 

 lum, however, we have such ten- 

 dencies as the insistence that all re- 

 search be set up so that it can be 

 treated statistically; that all results 

 be subject to statistical tests; that 

 a complicated technique be used 

 when a simpler one would suffice; 

 and that statistics be relied upon 

 for thinking and devising questions 

 that can come only from the re- 

 search worker. Such excessive use 

 of statistics is like using a lamppost 

 for support rather than for illumi- 

 nation. 



You Don't Have to Be an Expert 



There is no reason why one who 

 does not read a foreign language 

 should have to forego using the 

 contents of an article written in 

 that language, although the trans- 

 lator may well be handicaj)])ed by 

 a lack of technical knowledge in 

 the field involved. 



Similarly, you need not forego 

 the use of statistics simply because 

 the time and effort required to 

 gain an intelligent understanding 

 of them is more than you wish to 

 invest. But it is true that the stat 

 istician you consult may :.lso be 

 handicapped by a lack of back- 

 ground in your field. He will ap- 

 ]:)reciate it if you consult him be- 

 fore you collect your data and 

 reach your conclusions, rather than 

 decide to analyze statistically data 

 that are neither suited for nor 

 capable of analysis or to use your 

 own interpretation regardless of 

 the statistical analysis. You may 

 be right, but you should not give 

 the statistician credit for it. 



% 



