23 



labor income groups. They thus have no chance to cancel each 

 other. In fact the largest minus labor income group is made up 

 almost wholly of large farms, the average of this group being the 

 th-rd largest in the series. The irregularity in the size averages of 

 Table II-B is thus due to the fact that the basis of sorting did not 

 eliminate all the causes other than the one we were investigating. 



It will be noticed that in Table II-A the basis of grouping is 

 a causal factor, that is, variation in size of farms causes a correspond- 

 ing variation in average labor income; while in Table II-B the 

 basis of grouping is a resultant, or effect, in this case of many causal 

 factors. Grouping on a cause gave a consistent result, because the 

 effects of other causes could be eliminated, while grouping on an 

 effect gave a result that is hard to interpret because influences not 

 under investigation could not be eliminated. This divergence is 

 the usual result of these two methods of grouping. The beginner 

 in farm management investigations should never group his farms 

 on the basis of an effect; the grouping should always be on the 

 basis of a cause, and the averaging is then done on an effect. 



When the relation between a cause and the combined effect of 

 several causes, including the one in question, is not complicated 

 by any causal relation of both the cause and the combined effect 

 to some other causal factor, the variation in the averages of the 

 various groups of a tabulation will be due entirely to the variations 

 of the cause under investigation, and the table will display the 

 true effect of this cause. 



If the effect under consideration is not the combined effect of 

 two or more causes, but is solely the effect of the one cause under 

 consideration, then it is a matter of no consequence which method 

 of grouping be used that of Table II-A or that of Table II-B. 

 Suppose, for instance, that size of farm were the sole determining 

 factor in the matter of labor income; in such a case the labor income 

 of every farm would be directly proportional to its size, and group- 

 ing on either size of farm or labor income would throw exactly the 

 same farms into each of the groups. But such simple relations 

 are seldom found in farm management survey data. 



To show how the influence of one causal factor may sometimes 

 be deduced from the combined effect of this factor and another 

 which can not be separated from it, let us consider the combined 



