554 



BELL SYSTEM TECHNICAL JOURNAL 



into five different magnitudes for use in this theoretical calculation. 

 As a measure of the variation from equal loads submitted to the sub- 

 groups, the standard deviation of the estimated subgroup loads carried 

 taken about their average A'alue was used in each case. To estimate 

 the load, "/," which would be carried by any subgroup to which the 



if) 0.0100 



O 0.0010 



0.0002 







1.2 



1.0 

 u. Q- Z 

 O g Q 0.8 



2 o ^ 0-6 

 Q in cc 



Z D < 0.4 

 - 10 > 



0.2 



20 22 



24 26 28 30 32 34 36 38 



AVERAGE LOAD IN SIMULTANEOUS CALLS 



Fig. 19 — Comparison of theoretical and observed graded losses, "State-W abash" 

 group. .V = 8, >' = 12, g = 5. 



average, "o," is submitted, Erlang's formula, as set down earlier in 

 this paper, was utilized in the following relationship: 



/ = a[l - 5(.v, a)], 



wherein, as before, B{x, a) connotes the proportion of the submitted 

 load which goes beyond the .v individual trunks. The reason for 

 working with the theoretical load carried by a subgroup rather than 

 the load submitted to it will be clear when it is recalled that all the 

 observed data available for comparison yielded the former values only. 

 We shall now examine briefly the result of assuming subgroup variations 



