RESULTS OF REGRESSION ANALYSIS 



Equation (4), or a modification of it, was fitted to surface 

 temperature observations taken by a bathj'thermograph in the 

 areas A, B, C, and E shown in figure 4. In addition it was fitted 

 to the large area shown extending from 30 to 49 N and seaward 

 about 650 miles. In the latter area the sea-surface temperature 

 observations used were made in the four shaded 1 -degree-longitude 

 strips (B, C, D, and E) shown in figure 4. The measurements 

 were treated as a single sample drawn from the large area. 



Equation (4) was developed as a model for the largest area. 

 Since the other areas cover smaller intervals of latitude and 

 longitude, the terms in equation (4) that involve the higher orders 

 of these variables are omitted. A 3-month overlap in time was 

 used in making the least-squares fit to control the behavior of the 

 fifth-degree polynomial. Thus, the data used covered an 18-month 

 period extending from 1 April of a given year to I October of the 

 following year, and the resulting equation was used to estimate 

 the temperature during the included fiscal year. 



37 



