by
lmh » Fri Sep 01, 2017 1:03 pm
Tom's misunderstanding is just a matter of it being difficult to show equations in a typical forum editor - wish we could!
Basically the "standard compliance" method expresses the absolute difference between the two measurements as a fraction of the average of the two measurements, and therefore makes no assumption about which measurement is "right".
The ((larger/smaller)*100)-100 method assumes that the smaller measurement is right, and therefore gives the over-cautious percentage difference. If, for example, the measurements are 1000 and 500, it gives an error of 100%, meaning that the higher differs from the lower by 100% of the lower's value.
While generally I like caution, I believe the best estimate of a real value is likely to be the average of all the measurements available, so the "standard compliance" method is the most attractive. It's analogous to stating the standard deviation centred on the mean.