SMITHSONIAN MISCELLANEOUS COLLECTIONS VOL. 'J'] 



r — 



2XY 



V2X'-2F' 



Where -Y represents the departures from the mean value of the 

 forecasted temperatures, Y the departures from the mean value of 

 the observed temperatures, and r the correlation coefficient. For 

 verification purposes there were employed only the temperature 

 observations from the Villa Ortuzar Observatory, Chacarita, Buenos 

 Aires, though it probably would have been better to have used a 

 mean of several stations around the River Plata. 



Table i. — The method for calculating the correlation coefficient 

 August 14-20, 1924 



Date 



Temperature 

 mean of day 



Predic. Observ. 



XY 



y2 



14 

 15 

 16 



17 

 18 



19 



Sum 

 Mean 



-44 



Log. 12 = 1.079181 

 33= 1.518514 



2.597695 



I . 298847 

 Log. 19= 1.278754 



9.979907 

 ?- = 0.955 



The first step was to obtain the mean of the days for the predicted 

 and observed values and these were then used to calculate the cor- 

 relation coefficient as shown in table i.^ 



The correlation coefficients will be found in tables 2, 3, 4, 5 and 6, 

 representing two and a half years' forecasts from June, 1922 to 



^ Objection may be taken by some to the use of separate means by Mr. Hox- 

 mark for predicted and observed temperatures in computing departures. Yet 

 he merely gives himself thereby the slight advantage that he puts his forecasts 

 on the true scale. The variations from day to daj'- are not affected. 



