CARD TRANSLATOR EXPERIMENTS 375 



ai'o interested in e^■alnaling the interaction of the varial)les on each other, 

 when such intei'action exists and is of interest. In a sense tliis interaction 

 measures the departure of the system y, X, Z from Hnearity. 



Once the basic designs have been selected and ai)propriately combined 

 to fit most etfici(Mitiy the recjuiremcMits of \\w, pr()])()sed (experiment, 

 and the vahu^s of the variables randomly assigned to th(> schematic 

 layout, a detailed experimental layout must be drawn up. This layout 

 must show concisely and cleai'ly each experimental unit and the makeup 

 of every basic element giving its assigned value of each variable. Explicit 

 dii'ections must be drawn up as to the order of selection of the elements 

 of the unit. It is generally advisable for simplicity to assign the elements 

 at random to the M possible consecutive order integers of the experi- 

 mental unit. 



Performance Sludy 



The first seven variables hsted in the Introduction are: 



1. The Bin in use, Bins. 



2. Position of Test card pair within the bin, Position. 



3. The use of 3 digit or 6 digit Cards, Code. 



4. Arrangement of Coded and Uncoded Cards, Runs. 



5. Load of Bins containing Test Cards, Load. 



6. Load of Bins not containing Test Cards, Idlers. 



7. Order of repeated measurement. Look. 



These constitute a system — that is, any or all can be varied at will 

 and hence a design involving all of them simultaneously can be sought. 



The test set can operate ten test cards; 5 with a 3 digit code and 5 

 with a 6 digit code. This immediately suggests 5 packages of two coded 

 card pairs, each pair containing a 3 and a 6 digit card. Five pairs can 

 also be handled neatly in 5 bins. The combination of the standard load 

 of 85 cards, with two overloads and two imderloads w^ould give a fair 

 evaluation of load criterion. Budget restrictions force the use of only a 

 limited number of coded cards, with blanks used to fill out the experi- 

 ment. Hence the type of card making up the load must be varied over 

 the loads. It was further found that 5 positions of test cards within the 

 bin covers the range of positions adequately. 



The pairs of 3 and 6 digit cards now are associated with the Graeco- 

 Latin Scjuare design with Columns identified with Bins; Rows with dis- 

 tribution of coded cards; or Runs; Latin letters with Load; and Greek 

 letters with position within the bin. Now if the position of the 3 digit 

 card is randomly assigned in the pairs, the design abs()rl)s the first five 



