Statistical distance for random variables
D. FALIE1,*
Affiliation
- Faculty of Electronics, Telecommunications and Information Technology, University Politehnica of Bucharest, Bd. Iuliu Maniu 1-3, Bucureşti 030791, Romania
Abstract
Two statistical distances for random variables, based on the standard deviation operator, and on the mean absolute deviation, are proposed. The numerical values obtained with these relations are the same if both random variables are Gaussian, and may differ in the other cases. A generalized mean deviation covariance and correlation coefficients are also defined. These new statistical operators can be used to detect the non-Gaussian random or a mixture of random variables. The generalized statistical operators are differently affected by outliers and from these on optimum operator can be choose. These new statistical operators have been used for gaze guidance evaluations and for scan path measurements.
Keywords
Statistical distance, Non-gaussianity, Outlier, Generalized correlation, Generalized covariance.
Citation
D. FALIE, Statistical distance for random variables, Optoelectronics and Advanced Materials - Rapid Communications, 4, 7, July 2010, pp.1028-1032 (2010).
Submitted at: March 6, 2010
Accepted at: July 14, 2010