XXXIV] 



Introduction 



First Variate A. 



lix 



Illustration (ii). Find r^ by mean contingency for the table on p. lx: 



The sum of the positive contingencies is 169'846, or we have mean contingency 

 ■v|r = -170, whence the diagram leads us to rj, = '480. The marginal frequencies are 

 the same as in Illustration (i). Thus we have 



r = i (-517 + -480) = -499. 



The table gives actually a true Gaussian distribution with correlation -500. 

 It will be seen from Illustrations (i) and (ii), that if the distribution be Gaussian, 

 even if the marginal frequencies are in fairly irregular groupings, ?-^ will be 

 reasonably close to the true contingency, and corrected as suggested above will 

 give excellent results. 



hi 



