APPENDIX II. 305 
The coefficient of correlation may be very small and yet 
correlation may be a certainty if the coefficient is more than 
six times the probable error. 
In the case we have been considering, of the relation 
between supply and price, the coefficient was - -440 (i.e., 
inverse correlation), but the probable error was only -079. 
As this is less than one-sixth of the coefficient, correlation 
may be said to exist. In the case of wheat yield and rainfall 
for Canterbury, correlation was not established, for the co- 
efficient was only + °250 and the probable error -1. As was 
explained in Chapter IV.* this was due to the fact that the 
rainfall may vary considerably from the average without 
greatly affecting the yield. Thus movements in the yield must 
have been due partly to other causes. This is shown by the 
relatively high probable error. 
Further information upon the whole problem of correlation 
may be sought in 
King : Hlements of Statistical Method. 
Bowley: Elements of Statistics. 
Elderton : Primer of Statistics. 
*See page 67. 
