HKLlMilLirV OF TUAFFK' M IvVSUKFMKNTS 



:^(;i 



Of more accuralch' usiiii>; tlic solid ciiiAc 



1() 



The (MTor of usiiiu; (Mjuatioii ('2) is seen to 1h' ii(\<;Ii,ii:il)l(' for most jjuiposcs 

 o\-(Mi when T I is less than 20. The pi'ohahilit y of an ()l)s(M-\'ation oc- 

 cuiiiiii witliiii a .^ivcii miinlH'r of slaiulai'd deviatioii.s is widely published 

 for the iioiinal curve. A few \'alues are y:i\-en below: 



0.6745 



1.44 



1.64 



2.00 



3.00 



Pi Probability of exceeding ± jw or ± zP' 



0.50 



n.S5 

 ().«)() 



0.9545 

 0.9973 



Fig. 2 is a plot for 40 obser\-ations of measured load vs carried load. 

 Each observation was made for a half hour period on a panel hne finder 

 group with switch counts made at the start and middle of the period. 



I t.o 



;o.8 



) 



• 0.6 

 . 0.5 

 iO.4 



VvVdN , COEFFICIENT OF VARIATION OF ERROR IN ESTIMATING A ONE 

 ERLANG CARRIED LOAD IN ONE OBSERVATION PERIOD 



Fi^;. 1 — Accurtifv of switch count estimate of load actually carried. 



