Tests of symmetry derived as components of pearson's phi-squared distance measure |
| |
Authors: | Rebecca B Rosenstein |
| |
Institution: | Department of Community Health , Box G, Brown University , Providence, RI 02912 |
| |
Abstract: | Rank statistics which arise as estimates of the first and third components of a frequency decomposition of Pearson's phi-squared distance measure, introduced by Eubank, LaRiccia, and Rosenstein (1987), are examined for their usefulness as tests of symmetry about a known median against various asymmetric alternatives, Pitman asymptotic relative efficiencies are used to compare the efficacies of the new statistics with classical test procedures, and empirical powers of the new tests are compared via simulation for a variety of asymmetric distributions. Statistics which arise from components based on Legendre polynomials are of particular interest. The first component is the familiar Wilcoxon signed rank statistic. The third component, which is a new statistic for this problem, exhibits a high level of sensitivity to a variety of asymmetric alternatives both asymptotically and in small sample studies. |
| |
Keywords: | tests of symmetry rank statistics components of phi-squared |
|
|