CARD TRANSLATOR EXPERIMENTS 373 



Tlic design solcctod for the first experiment of this study combined 

 two basic desijiii types, the Latin S(iuafe,- and the Factorial Design.- 

 A biief discussion of (>acli is in ordei' at this point. 



The Latin S(iuare is peculiarly well suited to engineering research 

 problems \vher(> many variables exist but the luimber of discrete levels 

 necessary to describe^ the variation of any one is small, where relatively 

 great precision is possil)le in measurements, and where the interaction of 

 the variables is not a factor of the experiment. A key design- for a 5x5 

 Latin Square is as follows: 



Column 



It will be ol)ser\-ed that each Latin letter falls once and only once in 

 every row and column. It is also true that if the logical subgroups .1 , 

 B, C, D, E are considered, each of these sums has each row and column 

 included once. If to the five row- values, column values and letters, 

 the five values of the first, second and third variables are associated 

 respectively, the variables represented by Row, Column, and Letter 

 can be evaluated independently of each other. 



On the Key Latin Square, another properly chosen sfiuare represented 

 by Greek Letters can be superimposed, and the five values of the fourth 

 variable assigned to these letters at random : 



