Abstract
A full bivariate generalization of the Anscombe-Tukey one degree of freedom for univariate non-additivity is presented, and is compared to other proposals. Higher dimensional extensions follow directly. In general, k(k+1)/2 degrees of freedom can be allocated to assessing non-additivity in k dimensions. Hence coordinate-wise additivity is necessary but not sufficient for multivariate additivity.
Similar content being viewed by others
References
Anscombe, F.J. (1961). Examination of residuals.Proc. Fourth Berk. Symp. Math. Statist. Prob. 1, 1–36.
McDonald, L.L. (1972). A multivariate extension of Tukey’s one degree of freedom for non-additivity.J. Amer. Statist. Assoc. 67, 674–675.
Millikan, G.A. and Graybill, F.A. (1970). Extensions of the general linear hypothesis.J. Amer. Statist. Assoc. 65, 797–808.
Rao, C.R. (1973). Linear Statistical Inference and Its Applications. John Wiley, New York.
Scheffé, H. (1959). The Analysis of Variance. John Wiley, New York.
Tukey, J.W. (1949). One degree of freedom for non-additivity.Biometrics 5, 232–242.
Tukey, J.W. (1955). Answer to query 113.Biometrics 11, 111–113.
Author information
Authors and Affiliations
Rights and permissions
About this article
Cite this article
Koziol, J.A. Multivariate tests for non-additivity. Statistical Papers 30, 27–37 (1989). https://doi.org/10.1007/BF02924306
Received:
Revised:
Issue Date:
DOI: https://doi.org/10.1007/BF02924306