SMITHSONIAN MISCELLANEOUS COLLECTIONS VOL. 68 



Table 2. — Correlation Factors Computed from Observed Values 



Days following solar observations 



(i) July 16 to Aug. 24, 1913 



(2) Sept. 25 to Nov. 9. 1913 



(3) June 12 to Aug. 28, 1914 



(4) July 30-Aug. 5 and Aug. 29-31, 1914. 



(5) Sept. I to Oct. 20, 1914 



(6) All observations, 1913 



(7) All observations, 1914 



(8) Mean of 1913 and 1914 



iProb. 

 1 error 

 i of 

 ! max. 



i 

 ±.09 



.00 ±.09 

 !±.09 



.45'±-r2 

 !±.io 



±.07 

 ±.07 

 ±•05 



In computing" the values under (3), certain large values which 

 were considered doubtful by Dr. Abbot and his associates were 

 omitted. The correlation factor was computed separately for these 

 large deviations, and found to be very high as seen by the results 

 under (4) where the maximum correlation is nearly seven times as 

 great as the probable error. In every case excepting the interval 

 September to October, 1914, a positive correlation was found with 

 the maximum three to five times as great as the probable error. The 

 maximum coefficient for all the observations 191 3 to 1914 in (8) 

 although not large, proves to be five times the probable error, so that, 

 according to accepted standards, there is proved to have been a posi- 

 tive correlation during these years between variations in solar radia- 

 tion as measured at Mount Wilson, in the United States, and varia- 

 tions in daily maxima of temperature at Pilar, in Argentina. The 

 maximum correlation follows one to two days after the corresponding 

 solar values. In this respect the retardation is analogous to other 

 solar efifects. 



The maximum temperature of the day lags two to three hours 

 behind the meridian passage of the sun, and the maximum tempera- 

 ture of the year lags about a month behind the time of greatest altitude 

 of the sun. In each case the lag is one-tenth to one-twelfth of the 

 length of the period and by analogy irregular fluctuations of 5 to 15 

 days between maxima would show a lag of one to two days. It is 

 also worthy of note that the largest variations (4, table 2) not only 

 showed the maximum correlation, but also the greatest lag. Even 

 up to the fifth day the correlation was positive and nearly three times 

 the probable error. The probable error was determined by the 

 formula : 



P.E.= 



v» 



(2) 



