Separating reliability and stability in test-retest correlation.