Skip to main content
. 2004 Jul 15;112(15):1527–1538. doi: 10.1289/ehp.6976

Figure A2-1. Scatter plot of corr(Y,Yr) versus Q2[Yr]. The near zero y-intercept of the line plotted through the Q2 values confirms that the base model (right) is not due to random statistical associations, by showing that changes in the structure of the data decrease the performance of the model. High R2 for nearly all of the model permutations illustrates the need for the Q2 diagnostic because correlation (R2) occurs under several conditions without good model performance.

Figure A2-1