382 THE BELL SYSTEM TECHNICAL JOURNAL, MARCH 1954 



Table IV — Card Translator Balanced versus Unbalanced Load 

 Test — Arrangement of Coded Cards in Bin I 



Note: Numbers preceded by # represent the test cards. Numl)ers in italics 

 represent quantity of non-test coded cards placed between two consecutive test 

 cards. 



IV. ANALYSIS OF THE DATA ' 



After cheeking the raw data for recording errors, they were analyzed 

 and reduced in three distinct steps as follows: 



1. Using the techniciues of the analysis of variance* each variable 

 and each measured interaction of several variables was tested as a pos- 

 sible assignable cause of variation. 



2. Using the components of variance analysis! on those variables 

 and combinations of variables found to be assignable causes in Step 1 

 their contributions to the overall variation was estimated, and 



3. After tabulating the arithmetic mean for each value of the variables 

 an upper and lower bounds were determined which estimates the allow- 

 ance, or probable range of means, for similar experiments and opera- 

 tions on card translators having the same residual a. The technique used 

 was proposed and formulated by J. W. Tukey.^ 



The analysis of variance tables were reduced to the summary in 

 Tables XII and XIII for the seven variable studied. 



Proceeding to the tabulation of the means and the calculation of the 

 allowances appropriate to these means, the data was reduced as in Table 

 XI\\ The magnitude of the several effects can now be noted. The effect 

 of Idler load is c^uite meaningful — as the load on the machine is in- 

 creased to the normal load of 85 we see a decrease in mean dropping time 

 from 3G.5 to 34.4 milliseconds. Also the slight increase for the overload 

 of 100 is not significant statistically or engineeringwise. At first glance the 

 results of Idler loads of 0, 50, 85 and 100 might seem to be inconsistent 

 with those of Operating Loads of 15, 50, 85, 100, and 105. The mean 

 dropping time of the Load of 15 (say) is the a\-erage dropping time of 



* See Appendix for Discussion and References 2, 4 and 5. 

 t See Appendix for Discussion and Reference 6. 



