92 



BELL SYSTEM TECHNICAL JOURNAL 



have been replotted using instead of each actual message use 

 interval the per cent, that the message use is to average message 

 use. The result of this statistical process is to make the spread 

 of the curves and the height of the peaks much more nearly- 

 uniform. 



MAIN STATION 

 MESSAGE RATE SERVICE 



100 I 



150 200 250 300 350 400 50 100 150 20O 250 300 350 



MESSAGE USE AS PER CENT OF AVERAGE MESSAGE USE 



NOTE I- MA « MESSAGE ALLOWANCE. A.MM.U. - AVERAGE MONTHLY MESSAGE USE 



Fig. 6 — Sample Distributions of Subscribers by Message Uses in Per Cent, of Average 



In some instances, the frequency curves plotted with each interval 

 expressed as per cent, of the average may be so similar for the different 

 groups that satisfactory normals may be derived from this setup alone 

 without including any other factor. This appears to be the case for 

 the distribution of P. B. X. subscribers by message use as illustrated 

 on Fig. 6. It is necessary, however, to test whether or not the full 

 effect of the average on the distribution has been eliminated. This 



