On the Behrens–Fisher problem from the spatial median point of view
| Autor: | F. Rublík, Ján Somorčík |
|---|---|
| Rok vydání: | 2014 |
| Předmět: |
Statistics and Probability
Multivariate statistics Rank (linear algebra) Applied Mathematics media_common.quotation_subject Sample (statistics) Behrens–Fisher problem Test (assessment) Robustness (computer science) Modeling and Simulation Statistics Point (geometry) Statistics Probability and Uncertainty Normality Mathematics media_common |
| Zdroj: | Journal of Statistical Computation and Simulation. 85:3080-3091 |
| ISSN: | 1563-5163 0094-9655 |
| DOI: | 10.1080/00949655.2014.951358 |
| Popis: | We present a non-parametric affine-invariant test for the multivariate Behrens–Fisher problem. The proposed method based on the spatial medians is asymptotic and does not require normality of the data. To improve its finite sample performance, we apply a correction of the type which was already used in a similar test based on trimmed means, however, our simulations show that in the case of heavy-tailed distributions our method performs better. Also in a simulation comparison with a recently published rank-based test our test yields satisfactory results. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|