]\gt ; The last figures in are not all exact , and , in certain cases , the errors , which .
are larger than were estimated by and Paschen , amount to 1/ 100 of an ngstrom unit .
In the investigation , however , where the limit is calculated from every pair of successive lines , ) mean must be very accurate if the errors in the lines are not systematic .
The Hicks formula shows that they can be represented by , ( 8 ) where never exceeds .
If be expanded The fourth term of this very convergent series is .
Since is about , this cannot exceed 3 .
even when .
If we take all the lines beyond , it cannot exceed , so that for all these lines ; 'Handbuch der Spectroscopie , ' vol. 6 , p. 891 .
* Trans. Intern .
Union Solar Research , Bonn , 1913 .
The resulting weighted mean is xhaustive treatment drinci aseri , treated ieyond m , gives The general mean is from which we see that if , the only bad values are the first and last .
Now the first pertains to the lines most readily measured , and we ust conclude that the strict applicability of the method has not commenced this stage , so that the first two lines at least should be weighted on this ( account .
A similar error of opposite sign does not appear in the second value , as it should if an error of measurement accounted for the discrepancy the first value , according to the method used for the calculation .
The divergencies in the first and last values practically balance , so that the general mean must be nearly accurate , so far as casual observational errors are concerned .
By attaching weights 1 , 2 , 3 , 4 , 5 , 6 to the first six values , roughly to the degree to which the formula is applicable , and a weight 1 to the last value , we shall obtain a mean nearly free from errors formula .
The result is , and this result must possess considerable accuracy , beyond , in fact , the accuracy of individual lines .
We may now consider the values of The limits are ( Diffuse ) , ( Sharp ) , Principal )

