oneself up by one's own bootstraps (with perhaps the bootstraps being broken 

 in this case). Thus all of the above analyses and comments serve only to sug- 

 gest the nature and source of the error and a possible way to remove it. What 

 was done did remove the error, so in this sense the analysis of the error was 

 correct. 



All of the numerical results obtained in the original analyses of the full 

 sets of stereo spot heights were kept in the tables along with, the preceding 

 figures in order that this report would be complete. They represent a wealth 

 of data which can be used for additional analysis and study. This report is 

 unique in that it is a study of a random process in a plane^ and the complete 

 set of original data and computations should be of value to geophysicists, 

 statisticians and physicists o 



Re-analysis of reduced areas 

 As stated above both spectral computations were carried out over again 

 for reduced sets of spot height data„ For Data Set 2 the area was area A 

 as defined before as bounded on the four corners by the points 20,0; 20,49; 89j0; 

 and 89,49. In what follows these 3500 nu,mber s will be called Data Set 2A„ 

 Similarly for Data Set ?s area C (consisting of 3600 points) bounded by 0,10; 

 0,49; 89, 10; and 89,49 will be called Data Set 3Co 



The covarian ce surfaces for the reduced data 

 The covariance surfaces for Data Sets 2A and 3C and the average of 



the values for the two data sets are shown in figures 11<,13, 11.14, and 11.15. 

 The patterns are better defined than they were for the surfaces given previously 



181 



