114 DENSITIES OF OXYGEN AND HYDROGEN. 



I have entered my three results separately in the table, in order to indicate the 

 probable errors of the seveial values. It is obvious, from a compai'ison of the nine 

 results, that we may compute a mean in almost any way which even caprice may 

 suggest, without materially altering the vSlue deduced. We may assign ecjual 

 weiirhts to all the values, or weights proportional to the numbers of observations in 

 each seiies, or according to our judgment of the merits of the processes employed, 

 or inversely as the probable errors, or invei-sely as the squares of the probable 

 eri'oi-s ; in each case, our mean diffei's from the value obtained in the last case, 



0=15.879, 



by less than a ten-thousandth of this quantity. Until further light is obtained 

 concerning the sources of eiror which, doubtless, affect all these experiments, this 

 value is the most probable which can be derived from the existing data. 



V 



I 



