New Zealand Statistical Association
2004 Conference

Submitted Talks

NZSA 2004 Home ] Abstract Submission ] Call for Papers ] Conference Program ] Conference Registration ] Conference Timetable ] Important Dates ] Invited Speaker ] Organising Committee ] Submitted Talks ] Travel & Accomodation ] VUW Campus Map ]

About Wellington ]

Dong Wang
SMCS, Victoria University of Wellington

A New Approach for Detecting Multivariate Outliers

This paper starts with a short review of previous work on detecting outliers in multivariate data. Some theoretical aspacts are briefly discussed and then a new method to find outliers in the multivariate points cloud is introduced. At the same time an improvement on this method is developed. To apply these two methods, we then give several examples including one which demonstrates that the new methods are valid for detecting outliers in multivariate data, and a simulation example from which we can see how the improvement can be said an improvement, and finally a practical example about signal discriminant problems in DNA sequence analysis is given.