THE INTERCONNECTION OF TELEPHONE SYSTEMS 553 



seeming improvement of the observed data over its position in Fig. 16 

 is reasonable since here we have sorted out and combined only hours 

 of like average loads while before all the busy hours, high and low, 

 of a given period were included. It should be noted that the abscissa 

 for the observations here is load carried while for theory it is load 

 submitted. The comparison error is doubtless negligible for losses of, 



0. 1000 



Fig. 



O 0.0010 



0.0002 





 0.5 



°§^°-' 

 UJ O - r. o 



Q CD (T 0.2 



Z 3 < 



- 10 > 0.1 



n' l I I I i_1j I 1 1 1- 



10 12 14 16 18 20 22 24 26 28 HO 



AVERAGE LOAD IN SIMULTANEOUS CALLS 



18_Comparison of theoretical and observed graded losses, "State-Dearborn" 

 group. X = 4, J = 16, g = 5. 



say, P = .03 or less. The data beyond this figure are rather too 

 meager for useful correction. ^^ 



Next the "Full Gain" formula was generalized by the author to 

 comprehend the submission of different averages to each of the various 

 subgroups. (See Appendix II.) Then, selecting typical loads, for 

 instance A = 18.00 for the State to Dearborn tests and A = 33.80 

 for the State to Wabash tests, they were each divided up arbitrarily 



12 The advisability of correction here, were the data plentiful, niay well be doubted 

 since it would necessitate an assumption as to the manner m which calls were held, 

 a procedure especially precarious at high losses. 



