320 



for tliis purpose so that the labour ciilailed for tliis calciilatioii was 

 relatively small. 



The values not yet given are : 



r„ zn 0.G70 r„ = 0.359 »-,, = 0.744 



/•„ = 0.360 r„ = 0.781 



r,, = 0.543 r,, = 0.134 



r,,=r 0.888 >•,„ = 0.848 



And the condition equation becomes : 

 X, = 0.140 a-,-0.069 a-, -f 0.624 .r,-0.101 .v, -\- 



+ 0.538 .r„ + 0.015 .e^ (7) 



It appears from (7) that the methods of computation followed in 

 this inquiry fails in this case in so far that, owing to the insufficient 

 distances between successive stations, negative coefficients now appear 

 in the equations. Obviously they are due to a mutual distribution 

 of common influence which must be considered as unreal and as a 

 mere arithmetical result. 



Equation (7), therefore, shows a great resemblance to the first of 



he equations (6); the coefficients are alternatively small or even 



negative and if we reduce the equation to one with three terms by 



an equal distribution of the odd over the even coefficients so that 



for example : 



0.069 4 101 

 coëff. .v., = 0.624 = 0.539, 



we find the following equation little different from (6) 

 .V, =0.113 .V, + 0.539 A-, + 0.495 x^ 



In eipiation (7) the prevailing influence of the stations Miilhausen 

 and Sylt is still more conspicuous than in the results of other groups. 



A calculation of the remaining equation and of partial c.c. would 

 in this case have no meaning. 



Taken as a whole equation (7) is to be considered as an im- 

 provement because the general correlation-coefficient is very large 

 namely 



7^ = 0.9953 



from which follows, for the calculation of one value, the probable 

 error : 



M' ^ 0.539 mm. 



