Hi all,
I am measuring the agreement between a movement sensor that estimates VO2 in humans during exercise and criterion VO2. My supervisor advised me (and has done so in several of his publications published in high impact journals) to plot the difference against the criterion measure and not the difference against the mean of the difference. He says this is to prevent confounding.

Can anyone enlighten me on the pro's and con's of the two methods? I've trawled bland and altmans many papers and also their critics but still find nothing to support plotting the difference against the criterion.