I am currently trying to analyze some pupillometry data. The device I use gives me X and Y measurements of the pupil area (from which I compute the total area via $A = \pi \times \left(\frac{d}{2}\right)^2$ ).
The X-Y plot of my data looks like this:
In gray you can see my raw data points. As you see, though this effect does not seem to strongly affect participant means, there is a slight deviation from the regression line in the upper right quadrant of the figure.
My question is: in such a situation, would $d=X$ or $d=\frac{X+Y}{2}$ give me a more reliable measurement?