A new symmetrical test of bivariate independence based on correlation of scores is proposed. The involved test statistic is a renovation of the probability plot correlation coefficient proposed by Filliben. We analyze strengths and weaknesses of the new coefficient by comparing its finite sample performance to that of Pearson’s product moment correlation and to those of Fisher-Yates, Spearman, Kendall and Gini rank correlations, in both the absence and presence of outliers. Experimental results for real and simulated data reveal that the proposed coefficient performs reasonably well in evaluating agreement between variables presumably contaminated by outliers.
A NEW SYMMETRICAL TEST OF BIVARIATE INDEPENDENCE
Agostino tarsitano;ilaria lucrezia amerise
2017-01-01
Abstract
A new symmetrical test of bivariate independence based on correlation of scores is proposed. The involved test statistic is a renovation of the probability plot correlation coefficient proposed by Filliben. We analyze strengths and weaknesses of the new coefficient by comparing its finite sample performance to that of Pearson’s product moment correlation and to those of Fisher-Yates, Spearman, Kendall and Gini rank correlations, in both the absence and presence of outliers. Experimental results for real and simulated data reveal that the proposed coefficient performs reasonably well in evaluating agreement between variables presumably contaminated by outliers.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.