1 2 J 4 5 • 7 J ^ ' 



Figure 2. — The Standards best matched to the original aicrve bracket the averZay curve. 



5. Substitute Y for Y/Yp in the final, simple linear model to be fitted to a 

 relevant data set (the original set is used here) by least squares: 



Y = Bq + Bi x2-7 



Then the estimated Bg, Bj -- or Bg, B^ -- will scale the transforms to the Y^, 



It is generally preferable to use a computer for transforming the Xj_ and applying 

 the least squares fitting process and, in this case, the fitted model is 



Y = 2.06 + (0.8932 X 10"^) X^-^, <. X ^ Xp 



The curve for this model is compared to the original graphed curve in figure 4. 

 Differences indicate the extent to which the hand-fitted curve failed to follow the 

 least squares path. We assume here that the least squares curve is the most desirable 

 alternative within the X-range of the data. 



■^sing logarithm tables for the X^*^ transforms and a desk calculator for the fitting 

 process, the more laborious hand calculations would be as shown in Appendix B. 



5 



