NO. 5 



SOLAR VARIATION AND FORECASTING ABBOT 



17 



The values marked " computed " are obtained as follows : Taking 

 the value of the probable error as 0.0046 calorie, table 25 of the 

 Smithsonian Physical Tables enables us to compute at once how 

 many departures there should be according to the theor\' of accidental 

 errors up to the limits ±0.0015, ±0.0030, and so on up, in a series 

 of 398 observations. By subtraction, we obtain the numbers between 

 intervals ±0.0015 and ±0.0030, between ±0.0030 and ±0.0055, 

 et cetera. These are the values given in the lines marked " com- 

 puted " in the table. 



It will readily be seen that there is a goodly number of consider- 

 able departures, even exceeding ± i per cent of the solar constant, 

 where the theor}- of accidental error indicates that should be none 

 at all. No less than 18 of these extra, unexpected, values exceed 



Table 2.— Second Series. Departures in ten thousandths of a calorie. 



± about I per cent. This tends to prove that short-period solar 

 variation cannot be denied a real standing in court. We shall treat 

 long-period changes separately. 



We are at less advantage when we take the second series, because 

 the probable error is now 0.0065 instead of 0.0046 calorie. Never- 

 theless, the conclusions are much the same, as shown in the table. 

 There were here 744 cases in all, of which 30 more than the expec- 

 tancy exhibit dei^artures exceeding ± about i per cent, and 4 exceed 

 ± about 2 per cent. 



Figure 10, based on table i, shows strikingly the excess of large 

 observed departures over those predicted by the theory of accidental 

 errors. 



Thesis (b). — -Consistent evidence of solar variation is found in 

 the results of both stations. 



In figure 11, I show monthly mean values of both stations for the 

 years 1920 to 1922, The curves cover one of the most interesting 



