89,0 



n- 



0,0, 0,1, = . 0,4 j 0,5, 



I 



89,5 



-r 



-r 



0,9 



point s 



0.10 . 



0,49 



71" 



89,9 I 89,10 .... 89,491 89,50 ...89, 54 89,55. .. 89,59 



AP.EA 

 C 

 3600 



0,50. 



0,54 0,55. .0, 59 

 I 



The results of the analysis are shown in Table 11,3. Area C in this set 

 of data had the smallest variance (a reduction of about 0.24[(mm) (xlOO)]) 

 over the various points. However there is no significant discrepancy with any 

 combination of areas at the 5 percent level. 



The above results show something definitely wrong with Data Set 2 and 

 suggest something wrong in Data Set 3, especially since some spectral esti- 

 mates are negative in Set 3. It was therefore decided to do the computations 

 over again on a reduced portion of the data. The conaputations were performed 

 on area A (with 3500 points) in Data Set 2, and on area C (with 3600 points) 

 in Data Set 3. Some badly needed degrees of freedom were sacrificed by this 

 procedures but the results were quite encouraging. For example, the covari- 

 ances actually became negative on the vertical axis of the covariance surface 

 of Data Set 2, and there were no negative values in the smoothed spectral 

 estimates for either data seto A discussion of the corrected computations 

 will folio Wo 



178 



