AUTHOR'S NOTE 



This study was undertaken as a result of the following 

 recommendation made in an NEL report published in 1960: ^' 



"Explore the utility of multiple regression 

 (response surfaces), or more complex 

 analyses of variance, in summarizing the 

 entire body of collected data, since the 

 manner in which the observations are col- 

 lected limits the amount of the available 

 data which can be used for comparisons at 

 different places and times." 



This recommendation resulted from a review of the manu- 

 script of NEL Report 965 by Dr. George W. Snedecor, then Con- 

 sultant in Statistics at NEL. The initial attempts to apply regres- 

 sion models to summarizing sea-surface temperature data were 

 undertaken jointly by Dr. Snedecor and the author. The original 

 intent was to publish these results as a coauthored report. With 

 the retirement of Dr. Snedecor a number of years ago this is not 

 now possible. However, the author wishes to especially acknowl- 

 edge Dr. Snedecor' s enthusiastic motivation, interest, and con- 

 tribution to the results presented in this report. 



See references at end of report. 



