362 



THE BELL SYSTEM TECHNICAL JOURNAL, MARCH 1952 



26 



V- 24 

 Z 

 D 

 O 

 O 22 



20 



^ 12 

 < 



£ 10 



z 

 o 



_l 

 fR 6 



2 4 6 8 10 12 14 16 18 20 22 24 26 



CARRIED LOAD IN ERLANGS ESTIMATED BY 60 SWITCH COUNTS 



Fig. 2 — Accuracy of switch count estimate of true average load. 



28 



This is compared with the average of switch counts made every 30 sec- 

 onds which has a relatively negligible error. The average holding time 

 per call for the group was 176 seconds. The accuracy of only two 

 counts is surprisingly good and the observations are seen to he satisfac- 

 torily between the 2o- limits. 



ERROR OF TRAFFIC IN A GIVEN PERIOD AS AN ESTIMATE OF THE SOURCE 

 LOAD 



The average traffic carried in two different periods but generated by 

 the same traffic source is subject to statistical variation. As a result, any 

 measurement of load, even if measurement errors are eliminated, is only 

 a sample of the wide range of traffic loads that might have been gen- 

 erated by the same source of traffic under identical circumstances. 



