ON THE SEICHES OF LOCH EARN. 



457 



own, and the results entered in a time-book, from which correction curves were plotted 

 for reducing all times to the standard of my own watch, the irregularities of which 

 were found to be much less than other sources of error. 



In the tables which follow I have grouped together observations made under similar 

 circumstances, and deduced from each a weighted mean, thinking this the most impartial 

 way of treating the observations, with a view both to obtain the best final result and to 

 exhibit the degree of uncertainty of the observations. I have not taken the trouble to 

 calculate the so-called "probable error" in each case, as I have no assured faith in this 

 method for gauging human ignorance. 



To facilitate reference to the original material in the archives of the Lake Survey, 

 I have given the date of each observation in the first column of the table. In the 

 second column is given the mean height of the lake-surface above an arbitrary zero on 

 the measuring-staff. In the last column are given either the limits of error, as 

 estimated from the measurement of the limnogram, or else the number of oscillations 

 used in the determination. The weight in taking the mean is assumed to be inversely 

 proportional to the estimated possible error, or directly proportional to the number of 

 oscillations counted. 



I. Observations with Index Limnog raphs. 



Date. 



Staff. 



T r 



Limits of 

 Error. 



1905. 



Feet. 



Minutes. 









Minutes. 

 ±•10 



June 1 7 



1-35 



14-42 



24 



1-55 



14-50 



+ •10 



26 



1-51 



14-52 



+ •10 



26 



1-51 



14-60 



±•07 



30 



1-50 



14-50 



±•06 



July 1 



1-40 



14-46 



+ •02 



8 



1-35 



14-45 



+ ■10 



10 



1-35 



14-51 



+ 05 



25 



1-15 



14 59 



±•07 



25 



1-15 



14-64 



+ •20 



Aug. 7 



2-20 



14-58 



±10 



22 



2-40 



14-34 



±•10 



Weighted mean T x = 14492. 



