A Density Distance Based Approach to Projection Pursuit Discriminant Analysis
In this paper a brief review of the main contributions to projection pursuit discriminant analysis is presented. A new procedure, based on distance measures between probability densities, is developed in order to obtain linear discriminant functions, which best separate different populations. In the two- group case, Matusita’s distance between the projected population density functions is adopted as projection index; while, in the multi-group case, a monotone transformation of Matusita’s affinity coefficient is employed to measure the separation among the marginal probability density functions of the different populations. Simulation studies stress the efficacy of the proposed method in comparison with classical parametric ones and with the projection pursuit based linear discriminant procedure developed by Posse.
Key wordsMatusita’s Affinity Coefficient Projection Pursuit Discriminant Analysis Matusita’s Distance Coefficient
Unable to display preview. Download preview PDF.
- Bhattacharyya, A. (1943). On a measure of divergence between two statistical populations defined by their probability distributions, Bulletin of the Calcutta Mathematical Society, 35, 99–109.Google Scholar
- Calò, D. & Montanari, A. (1997). An empirical discrimination algorithm based on projection pursuit density estimation, Book of Short Papers, Classification and Data Analysis, Meeting of the Classification Group of Società Italiana di Statistica, Pescara, 3-4 luglio, 65–68.Google Scholar
- Chen, Z. & Muirhead, R.J. (1994). A comparison of robust linear discriminant procedures using projection pursuit methods, in Multivariate Analysis and its Applications, IMS Lecture Notes-Monograph Series, Hayward, 24, 163–176.Google Scholar
- Friedman, J.H. (1985). Classification and multiple regression through projection pursuit, Technical Report 12, Laboratory for Computational Statistics, Department of Statistics, Stanford University.Google Scholar
- Henry, D.H. (1983). Multiplicative models in projection pursuit, Ph. D. Thesis, Stanford Linear Accelerator Center, Stanford University.Google Scholar
- Huber, P.J. (1990). Algorithms for projection pursuit, Technical Report 3, Department of Mathematics, M.I.T. Cambridge.Google Scholar
- Krzanowski, W.J. (1988). Principles of multivariate analysis, Oxford Science Publication, New York, 359–360.Google Scholar
- Montanari, A. & Lizzani, L. (1996). Projection pursuit e scelta delle variabili, Atti della XXXVIII Riunione Scientifica della Società Italiana di Statistica, 2, 591–598.Google Scholar
- Roosen, C.B. & Hastie, T.J. (1993). Logistic response projection pursuit regression, Statistics and Data Analysis Research Department, AT&T Bell Laboratories, Doc. BL011214-930806-09TM.Google Scholar
- Sheather, S.J. & Jones, M.C. (1991). A reliable data-based bandwidth selection method for kernel density estimation, Journal of the Royal Statistical Society, Series B, 93, 683–690.Google Scholar
- Silverman, B.W. (1986). Density estimation for statistics and data analysis, Chapman and Hall, London.Google Scholar