RELIABILITY OF IIOLDIXG TIML M EASl'RKMK.XTS 369 



Tn Table T and on Fii;. 2 arc shown the liolding time averages for \M^ hours 

 observed at various times of the (hiy over a [)eriod of .^ months. At lirst 

 glance these appear to fall in two rather distinct groups, those before noon 

 and those after noon. If the 115 hours before noon be considered as 

 defining a homogeneous group, could those holding time averages found in 

 the afternoon be reasonably considered as coming from the same universe? 

 We first find the average holding time of the 115 forenoon hours to be 

 143.5 seconds. Since these hours averaged about n = 390 calls each, the 

 standard deviation of the means a-, should, by theory, be closely 



(Ti 



Vn \/390 



^ = 7.29. 



12-1 1-2 



HOUR OF THE DAY 



Fig. 2 — Da>' to da>- holding time averages by hours of the day, 135 hours, Newark 



The standard deviation observed is 9.26, some 27% higher, which, however, 



agrees with the observation made in the previous paragraph. On the 



hypothesis that the universe of 115 early hours has the parameters of / = 



143.5 and a — 7.29, we see that the observations for each of these three 



clock hours could readily have occurred. The deviations of their averages 



from 143.5 are 1.9, .17 and 2.5 seconds, respectively, and according to the- 



/ 7.29 

 orv the corresponding standard errors in these averages are 1 .168 I — .__ 



\ \/39 



1.111 ( = '!_ ), and 1.270 ( = — ^^_ ]. All the deviations are well within 



\ V43/ \ V3>.W 



two times the standard error of the assumed mean of the holding time uni- 

 verse. The remaining 20 observations from noon on, however, average 154.6 

 seconds, and if they could reasonably have come from the hypothesized uni- 

 verse, this figure should not differ from 143.5 by more than, say, three 



/ 7 29\ 



times the standard error 1.630 ( = — -=_ ). Actuallv the dil!"erence is more 



V2O/ 



