46 WHEAT PRODUCTION IN NEW ZEALAND 
The yield per acre for the whole province of Canterbury 
had to be taken because no separate continuous statistics 
for any district near the meteorological station are avail- 
able. Probably this is the weak spot in the investigation ; 
for we should have composite rainfall statistics at several 
stations in Canterbury, when the yield refers to the whole 
province. But with the available statistics of rainfall 
from May to November and average yield per acre, I 
have obtained over the years 1871-1918 a co-efficient of 
correlation* of + -251. As the probable error is as much 
as -1 correlation does not seem to be established in any 
degree. 
The defects in the information as shown above, must 
be kept in mind, as also the fact that wheat has a wide 
range of rainfall, so that a country where the average 
rainfall is about in the middle of this range, as it is 
in New Zealand, may show wide fluctuations in rainfall, 
and get comparatively small fluctuations in yield. 
From the Agricultural College and Experimental 
Farm at Lincoln I have been able to get statistics of 
yield and rainfall there, for the past 12 years (1906-17). 
The statistics are shown in graph No. 4, but the period 
is too short to afford a conclusion of importance. The 
continuance of these records will provide useful infor- 
mation in regard to the connection between yield and 
rainfall in Canterbury. It may be observed, however, 
that the curve of yield per acre moves in sympathy with 
the curve of ‘‘useful’’ rainfall, though the range of 
fluctuation is greater. This is suggestive of a close 
connection between the two series. 
It is interesting to compare similar statistics in Aus- 
tralia. A graph showing yield per acre for Victoria and 
composite winter rainfall (May-October) shows a vita] 
connection between yield and rainfall. This information 
was supplied to me by the Department of Agriculture 
for Victoria, and the result in the form of a graph is 
highly suggestive of a vital connection between yield 
*Correlation is dealt with again in Chapter VI., and is fully 
explained in Appendix IT. 
