st: RE: Interpretation of the Bland-Altman method on Stata

["Seed, Paul" <paul.seed@kcl.ac.uk>]

Dear Emily, 

-baplot- clearly works with logged values; or at least produces numbers rather than error messages. 
When I said it was not designed for logged values, I meant that the output is not clear, 
and further calculations are needed to make sense of them.

It is clear that you used logs base e, not base 10.  
-baplot- treats the first variable as the new method and the second as the reference; 
so you might want to reverse the order.

The mean difference is based on 
               energy_log_r - energy_log_q
               = log( energypd_r)-  log( energypd_q)
               = log( energypd_r/energypd_q)
So exp(mean difference) =  energypd_r/energypd_q
Simple arithmetic converts this ratio to a % difference:

exp(-0.420) = 0.657
Because you are using logs this is the ratio of the geometric means.
The value using method r is only 66% of that found using method q.
A 34% reduction.