( 32 ) 



c. We can assume tliat the evidence is equally good for the three 

 periods, and \\ ill be enhanced l»y continued arrangement. 



Of course o\\\\ an actual continuation of arrangement for another 

 sei-ies of twenty years will enalde us to answer these questions. 

 The only test which at present can be applied is to form two or 

 more groups as has lieen done abo\c for the arrangement according 

 to 25.8 da vs. 



So far as this ci-iterion allowN a conelusion to be drawn, it appears 

 from this i-esnlt that the evitlence for real existence of the ()eriods 

 25.70 and 25.87 is considerably less than of the period 25.80. 



In the latter case the arguments for three groups did not differ 

 more than 25°, against differences of resp. 88° and 95" for periods 

 of 25.70 and 25.87 days. The probabilities (»f mere chance, therefore, 

 are, takuig 30^ and 90° : 



1 , 1 

 and — 



144 1(5 

 i. e. uiore than 8 times as great. If we take also into account that 

 the aiuplitiides of the three groups are accordant for 25.80 and 

 widely different for the adventitious periods. ^^ c can estimate the 

 proi)ability of chance at 10 times as great. 



The computation of the probable error (incertitude) of the result 

 for each group also gives an indication for this greater probability, 

 but not in the same degree. 



Amplitude. Probal>le error. 

 25.80 1.76 mm. 0.8J0 



24.70 1.76 „ 0.916 



25.87 1.41 „ 0.830 



7. If we apply, in so far as possible, the different criteria to the 

 data published by Prof .Schuster concerning daily means of magnetic 

 declination for Greenwich, arranged according to 26 and 27 days, 

 we find for the sums of groups, each of which contains i-csp. 14 and 

 13.5 rows. 



