Some Evidence on the Largest Squared Correlation Coefficient from Several Samples