The variance of area A was 0.50[{n'Lm) x 100] less than the \'ariance 

 of the total area. The variance of area C was over twice as large as that of 

 area A. The degrees of freedom actually used in the F test were less than 

 the computed degrees of freedom, and yet at the 5 percent significance level 

 the hypothesis that the sample of points from area C is from the same popula- 

 tion as the sannple of points from area A must be rejected. 



The grid of points for the numerical analysis must be rectangular. Areas 

 B and C, D and Ci and B, C, and D were combined, and their combined 

 variances were tested against area A. In all combinations, the areas could 

 be rejected at the 5 percent level. Moreover the lower confidence bounds on 

 the variances of area C, areas B+C, C+D, and B + C + D were all greater 

 than the upper confidence bound on area A. 



Figure 11.12 shows a comparison of the probability histograms (number 

 of points in class interval divided by total number of points) from areas B, 

 C, and D, with the probability histograna from area A. Ai^ea C is quite a 

 bit different from area A. Note also that the histogram for area A appears 

 to be nornaalo 



The spot heights for Data Set 3 were analyzed in a similar way. A 



study of the contours suggested that a tendency toward vertical instead of 



diagonal crest orientation existed on both edges of the area of analysis and 



the points in Data Set 3 were broken up into five areas as indicated 



below. 



176 



