Before the analysis of the corrected results is mades a discussion of 

 what went wrong with the original results is needed. The basic source of the 

 difficulty can be traced back to a statement made in Part 6„ The photographs 

 were taken with reconnaissance type film instead of the more dinciensionally 

 stable topographic base film. The film magazines used in the cameras were 

 labeled to contain the correct film but they had actually been loaded with the 

 wrong film. Such a mistake would not be detectable until after the film had 

 been developed. This dimensionally unstable film then underwent differential 

 changes in areas (that is, small areas of the filixi shrank by greater amounts 

 than others) which introduced a complicated error pattern in the spot height 

 data. Fortunately most of the error (but possibly not all) appears to have 

 been concentrated on the edges of the areas analyzed. 



The question might be asked as to why the errors in the original leveled 

 spot heights were not detected prior to making the laborious computations of 

 the covariances and spectra given above. A close comparison of figures 11. 1 

 and 11,3 suggests, since hindsight is always better than foresight, that the 

 error in the spot heights might have been detectable simply on a comparison 

 basis. To be really sure, however, computations similar to the ones given 

 above would have had to have been made, and they could not have been naade 

 without a knowledge of the effective number of degrees of freedom of the sub- 

 samples. This effective nunaber of degrees of freedom was estimated from 

 the incorrect spectra. The use of theories valid for correct data on incor- 

 rect data to show that the data are incorrect is quite similar to pulling 



180 



