A METHOD OF GRAPHICAL ANALYSIS 



93 



may be done on a detail basis by plotting a series of charts showing 

 the relation between the absolute amount of the average and the per 

 cent, of cases falling within a given message use interval (expressed as 

 per cent, of the average). On a cumulative basis the per cent, of 

 cases falling below a given per cent, of the average is plotted against 

 the average. For example: 



AVERAGE MONTHLY MESSAGE USE 

 Fig. 7 — Preliminary Chart of Cumulative Series 



Fig. 7 shows the relationship between average message use and 

 the per cent, of subscribers using less than 100 per cent, of the 

 average message use for a given service classification. Each 

 plotted point represents the reading from the cumulative curve for 

 a different exchange. It is evident that the two factors vary 

 together. 



Curves similar in type to that shown in Fig. 7 are constructed on 

 each of the charts of the detail and cumulative series. The curves 

 of each series are smoothed by cross-sectioning and developed into 

 consistent and reasonable families. 



In connection with the smoothing of the cumulative series a method 

 described below has been found useful. This method can be used 

 with any setup of three variables but is simplest in setups of a cumula- 

 tive type which have no maxima or minima within the limits of the 

 curves. To simplify the explanation of the method the cumulative 

 distribution of certain subscribers by message use is referred to as an 

 example. 



Fig. 8 shows preliminary curves representing the relationship 

 between average message use and the per cent, of subscribers using 

 less than the various per cents, of the average message use from 10 

 per cent, to 500 per cent, of the average. These curves are derived 

 from a series of charts similar to Fig. 7 for different message uses. 

 Cross sections of the family of curves of Fig. 8 give a series of cumu- 



