of the Barometer near Edinburgh. 157 



draw more delicate conclusions, such as the influence of the sea- 

 sons, would be vain ; and, finally, I was led to distrust the ac- 

 curacy of even the general result. Under this impression, I 

 resolved to repeat the whole reductions upon a plan which I had 

 previously employed with success, and which has fully answered 

 my expectations. As the great object was to avoid the inequality 

 of absolute height, which might vary indefinitely on different 

 days, and preserve only the differences corresponding to different 

 hours, I selected one time of day at which the barometer was 

 most regularly observed, and, under the columns corresponding 

 to the other hours, inserted the differences as they were -f- or 

 of the barometer and attached thermometer, from the heights at 

 the standard hour inserted in the first column, and the means 

 for each month of the principal column and the subsidiary ones 

 being taken, the corrections (total or partial respectively) for 

 temperature were applied, from which mean results a new set of 

 absolute heights might be deduced. By far the greater devia- 

 tions from the mean, on the account above alluded to, were thus 

 avoided, and though particular omissions no doubt affected the 

 partial means, the errors were greatly smaller in amount. As this 

 simple artifice is not, I believe, generally adopted, and as it seems 

 greatly preferable to the employment of interpolation which 

 must frequently be guessed at, and always open a door to the 

 practice of admitting hypothetical numbers on the same footing 

 with observed ones, I think it worth while to give a practical 

 example of the mode I pursued, and the forms of my Tables. 

 The normal hour employed was 10 p. M., at which time the ob- 

 servations appear to have been been most regularly made ; and 

 it is surprising how few observations I was obliged to discard 

 for want of this standard of comparison. I have selected the 

 Tables for June 1828, in which 18 observations out of 150 were 

 wanting, the first reduction of which, by the method of simple 

 means, gives the oscillation from 10 A.M. to 4 p. M. in the wrong 

 direction, which is again restored by the application of the 

 second method I employed, shewn on the opposite page. 



VOL. XII. PART I. X 



