r2 value in regression depends on the values of the independent variable one choses.
For example, for exactly the same input T and output Y of an experiment (e.g. set a temperature t and measure y(t) of the system), r2 changes between the regressions with a narrower and wider range of T.
Thus, one cannot compare r2 of a relationship between X and a variable Y with different ranges, or r2 of a relationship 1 (X and Y) and relationship 2 (X and Z).
On the other hand, you can get correlation coefficients r_a (between variable X and Y) and r_b (between variable X and Z) and compare r_a and r_b, to tell which Y and Z are more correlated with X.
