CORRELATION. 



201 



Til"? 34 months' observations at South Orkneys give a correlation coefScient with 

 McMurdo Sound of +'15 with a probable error of -12 and the 19 months' at South Georgia 

 a correlation coefficient of -|-'21 with a probable error of -15. These coefficients are small 

 but positive. 



Collecting our results we have the following correlation coefficients between the monthly 

 pressure in McMurdo Sound and at five other Antarctic stations. 



Table 120. 



MontMi/ pressure correlation coefficienls. 



The absolute value of these coefficients means very little and they are not directly com- 

 parable with one another as they refer to different periods, but the fact that they are all 

 positive and are generally considerably larger than the probable error leaves little doubt that 

 the pressure tends to increase and decrease simultaneously over the whole Antarctic Continent 

 and the adjacent seas. 



Pressure correlation between the Antarctic and surrounding places. 



We can now turn our attention to the correlation between the Antarctic pressure and 

 the pressure at other places, mainly in the southern hemisphere, outside the Antarctic. 



If it had been possible, it would have been best to correlate the mean pressure of several 

 places in the Antarctic with the pressure at other stations. But for this purpose we have 

 only the twelve months' simultaneous data from Hut Point, the Gauss Station and Snow 

 Hill which would be insufficient for a reHable correlation. Above every thing a large number 

 of observations are necessary in order to get a reliable correlation coefficient. It was therefore 

 decided to take the four years' data which are available for McMurdo Sound, namely, February 

 1902 to January 1904 and January 1911 to December 1912, as being representative of the 

 whole Antarctic. 



In the Antarctic the changes of pressure from month to month are large in comparison 

 with the yearly variation, but this is not the case with stations nearer the tropics. It is 

 therefore necessary in order to get a true correlation between variations of pressure to eli- 

 minate the yearly variation from all the data used. 



It is also important to realise that the departures to be compared must be departures 

 from the same datum at all stations. Thus it is not desirable to compare the departures 

 for the Antarctic which are the departures from the mean of a certain four years with 

 the departures for another station from a mean derived from a long series of observations. 



In view of these considerations it was decided to follow a common procedure with the 

 data from all stations having the necessary records. The mean monthly pressures for the 

 forty-eight months for which data are available at McMurdo Sound were tabulated for each 



2G 



