432 
DE. T. E. EOBINSON ON THE SEDUCTION OE ANEMOGEAMS. 
statements, that casual irregularities are eliminated from the mean of a large number of 
observations. Still I think that a truer result might be obtained by omitting extremely 
aberrant observations ; but it becomes a question to what extent this should be done. 
I think all maybe rejected which exceed four times the largest probable error; in other 
words, whose probability is less than 0-0228. This is for W— W' all above 15. The 
number of these is 58, and the results after their exclusion are given in Table XI. 
Table XI. 
Month. 
Current. 
No. 
PE 
Weight. 
CxW. 
January 
0-1498 
386 
±1-592 
1*321 
0*1979 
The mean according to the 
weights is C"= 0*0559- 
February 
0-2483 
366 
±1*607 
1*317 
0*3204 
March 
-0-0423 
406 
±1*726 
1*266 
— 0*0535 
April 
-0-0590 
381 
±1*657 
1*289 
— 0*0760 
0*0024 
May 
0*0010 
419 
±1*262 
2*438 
June 
0-0718 
402 
+ 1*204 
2*496 
0-1753 
0*0608 
2360 
±1*443 10*127 
0*5665 
The probable errors are less, and the weights greater than in the other case, so that 
C" is probably a better value than C'. 
It is possible that this mode of proceeding might give the horary changes of the 
coordinates more correctly than the simple comparison of the numbers in Table III. ; 
but the labour of computation would be much greater. 
