CORRELATION. 



Table 121 — condd. 



203 



The map shows that encircling the Antarctic is a region in which all the correlation 

 coefficients are negative. It is impossible to define this region with exactitude, but as the 

 maximum coefficients both in the American and Australian regions lie near to latitude 40° S., 

 this latitude is probably near the axis of the zone, and the region is probably fairly accurate- 

 ly shown by the shaded area on the map. This region includes the large areas of high 

 pressure which encircle the globe near to latitude 30° S., and the most simple explanation 

 of the pressure relationship under discussion is that there is a sea-saw of pressure between 

 the Antarctic and the belt of anticyclones over the Southern Ocean. 



The largest correlation coefficients are --49 at Wellington, New Zealand, and —'.54 at 

 Cordoba in South America. The probable error of these coefficients is as low as -07, therefore 

 the coefficients are seven times greater than their probable errors. Thus there can be little doubt 

 that a month of high pressure over the Antarctic is accompanied by a month of low pressure 

 in latitude 40° S. and adjacent regions. 



It is interesting to notice that there is considerable difference between the correlation 

 coefficients in the two periods 1902-03 and 1911-12, but the differences are not random. 

 In the last column of table 121 the change in the coefficients from 1902-03 to 1911-12 has 

 been entered for each station. Over the whole of the Australian region including Samoa and 

 Fiji the change was negative, while with two exceptions the change was positive over the 

 American region. It would appear as if the interchange in the air during 1902-03 was mainly 

 between the Antarctic and the American region, while in 1911-12 the change was between 

 the Antarctic and the Australian region. 



The positive coefficients in the north of Australia are interesting. One would have ex- 

 pected the coefficients to have become less and less negative as one receded from the region 

 of maximum correlation. That they should become positive and decidedly positive, is very 

 remarkable. It will also be noticed that in 1902-03 the region of positive correlation included 

 the whole of the Australian Continent except the extreme south-east corner, and in the 



