2 SMITHSONIAN MISCELLANEOUS COLLECTIONS VOL. 7 1 



solar values, the values of solar radiation observed in Calama, Chile, 

 in 1918, from July to September, were first correlated with the tem- 

 perature in Buenos Aires in the usual way, the correlation being 

 made for the same day as the observed solar value and for each 

 succeeding day up to lo days later. These values were kindly 

 furnished me in advance of publication by the Secretary of the 

 Smithsonian Institution. 



The correlation formulas used were : 



,- = -.^L_-^. ( I ) r = o. 674 (i-~-:'-^ ( 2 ) 



in which the values of x are the deviations from the mean solar value, 

 and y, deviations from normal temperature in Buenos Aires ; n is the 

 number of observations, and e is the probable error when n is large. 

 When the means of x or y differed from zero during the interval 

 covered, a correction was applied. 



Next, all values near the mean were suppressed and only values 

 used which differed from the mean as much as 0.020, or about i per 

 cent of the mean radiation value. 



From these latter values a new correlation factor, r', was com- 

 puted. 



The results are given below, the figures following r showing the 

 correlations when all the values were taken, and those following r' 

 showing the correlations with the larger values of solar radiation. 



Table I. — Correlation Between Temperature in Buenos Aires and Solar 

 Radiation, July to September, igiS 



Days follow solar variations 



0123456 789 10 



r= — 16 — 20 — 21 — 14 -I-02 +09 +22 +38 +35 -\-23 +14 

 ;■'= —47 —34 —47 —65 -1-08 +28 +58 -1-64 +68 +44 +32 



These data indicate very clearly that the correlation is not a linear 

 function, not only because the correlation values are larger when the 

 data are separated into two classes, but because the maxima and 

 minima occur later. In the mean of all, a maximum negative correla- 

 tion occurs on the second day and a maximum positive on the seventh 

 day following the solar occurrences, while after large values the 

 maximum negative value occurs on the third day and the maximum 

 positive occurs on the eighth. With the data for 1913 the correlation 



