Abstract
Various distances between distributions and between densities are considered. The corresponding goodness-of-fit tests derived from them are examined for their abilities to detect multimodal alternatives. It is found that many well known techniques fail to detect such alternatives, while others do better in terms of their power results. These are mainly the tests derived from the variational metric which are based on spacings and gaps.
Original language | English |
---|---|
Pages (from-to) | 421-439 |
Number of pages | 19 |
Journal | Metrika |
Volume | 42 |
Issue number | 1 |
DOIs | |
State | Published - Dec 1995 |
Keywords
- EDF
- Goodness-of-fit
- Multimodal Distributions
- Spacings
- Variational Metric
ASJC Scopus subject areas
- Statistics and Probability
- Statistics, Probability and Uncertainty