Extended critical vawes of the multivariate extreme deviate test for detecting a single spurious observation |
| |
| Authors: | Linda W. Jennings Dean M. Young |
| |
| Affiliation: | 1. Academic Computing Services Division , The University of Texas Southwestern Medical Center , Dalla , TX , 75235;2. Department of Information Systems , Baylor University , Waco , TX , 76798 |
| |
| Abstract: | The Institute of Mathematical Statistics has published a table of critical values for the multivariate extreme deviate test. However, the critical values, derived by a Monte Carlo simulation, are given for only the dimensions 2 through 5. We present new critical values for the dimensions 6 through 10, 12, 15, and 20. The results are presented in both table and graphical form. All critical values for the test statistic have been generated by a Monte Carlo simulation using 10,000 observations per case. An example is presented using the new critical values. |
| |
| Keywords: | Monte Carlo simulation high-dimensional outlier detection smoothing |
|
|
