The scaling procedure for sets 1 and 2 is shown in the example that follows. Set 

 3, because it involves reciprocals and because the latter are indeterminant at X = 0, 

 requires that Yp be determined within a slightly abbreviated X-range as explained in 

 the last paragraph under the section labeled "An Example." 



For the sake of efficiency, then, it is 

 scaled first for comparison to sets 1 and 2— 

 set 3 only when no suitably similar form can 



suggested that the analyst's curve be 



; the curve will be scaled for comparison to 



be found in sets 1 and 2. 



Where ^the analyst's curve lies either partly or wholly in some quadrant other than 

 "upper right" (i.e., not oriented as the curves above), adjust either the X- or Y-scale, 

 or both, to put the curve on the same basis as the standards before applying Matchacurve 

 --see Appendix A. 



AN EXAMPLE 



This example applies as shown to sets 1 and 2. 



Let's say that an analyst wants to generate an equation for Y from the data set 

 plotted in figure 1. From previous knowledge of the relation between X and Y, he 

 expects the curve to be concave upward with larger values of X. The apparent data trend 

 supports his expectation and he hand-fits such a curve through the data, also shown in 



-^his scaling procedure is identical to that employed for the 1970 companion paper 

 for sigmoids, Matchacurve- 1 , so the same scaled curve can be compared to the sigmoid 

 standards, if pertinent. 



