Abstract
In multivariate location problems, the sample mean is most widely used, having various advantages. It is, however, very sensitive to outlying observations and inefficient for data from heavy tailed distributions. In this situation, the spatial median is more robust than the sample mean and could be a reasonable alternative. We reviewed several spatial median based testing methods for multivariate location and compared their significance level and power through Monte Carlo simulations. The results show that bootstrap method is efficient for the estimation of the covariance matrix of the sample spatial median. We also proposed bootstrap simultaneous confidence intervals based on the spatial median for multiple comparisons in the multi-sample case.
| Original language | English |
|---|---|
| Pages (from-to) | 2123-2133 |
| Number of pages | 11 |
| Journal | Communications in Statistics: Simulation and Computation |
| Volume | 38 |
| Issue number | 10 |
| DOIs | |
| Publication status | Published - 2009 Nov 1 |
| Externally published | Yes |
Fingerprint
Keywords
- Bootstrap
- Multivariate location
- Simultaneous confidence interval
- Spatial median
ASJC Scopus subject areas
- Modelling and Simulation
- Statistics and Probability
Cite this
Bootstrapping spatial median for location problems. / Jhun, Myoungshic; Shin, Seung Jun.
In: Communications in Statistics: Simulation and Computation, Vol. 38, No. 10, 01.11.2009, p. 2123-2133.Research output: Contribution to journal › Article
}
TY - JOUR
T1 - Bootstrapping spatial median for location problems
AU - Jhun, Myoungshic
AU - Shin, Seung Jun
PY - 2009/11/1
Y1 - 2009/11/1
N2 - In multivariate location problems, the sample mean is most widely used, having various advantages. It is, however, very sensitive to outlying observations and inefficient for data from heavy tailed distributions. In this situation, the spatial median is more robust than the sample mean and could be a reasonable alternative. We reviewed several spatial median based testing methods for multivariate location and compared their significance level and power through Monte Carlo simulations. The results show that bootstrap method is efficient for the estimation of the covariance matrix of the sample spatial median. We also proposed bootstrap simultaneous confidence intervals based on the spatial median for multiple comparisons in the multi-sample case.
AB - In multivariate location problems, the sample mean is most widely used, having various advantages. It is, however, very sensitive to outlying observations and inefficient for data from heavy tailed distributions. In this situation, the spatial median is more robust than the sample mean and could be a reasonable alternative. We reviewed several spatial median based testing methods for multivariate location and compared their significance level and power through Monte Carlo simulations. The results show that bootstrap method is efficient for the estimation of the covariance matrix of the sample spatial median. We also proposed bootstrap simultaneous confidence intervals based on the spatial median for multiple comparisons in the multi-sample case.
KW - Bootstrap
KW - Multivariate location
KW - Simultaneous confidence interval
KW - Spatial median
UR - http://www.scopus.com/inward/record.url?scp=84901387681&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=84901387681&partnerID=8YFLogxK
U2 - 10.1080/03610910903249528
DO - 10.1080/03610910903249528
M3 - Article
AN - SCOPUS:84901387681
VL - 38
SP - 2123
EP - 2133
JO - Communications in Statistics Part B: Simulation and Computation
JF - Communications in Statistics Part B: Simulation and Computation
SN - 0361-0918
IS - 10
ER -