An evaluation of bootstrap methods for outlier detection in least squares regression
Outlier detection is a critical part of data analysis, and the use of Studentized residuals from regression models fit using least squares is a very common approach to identifying discordant observations in linear regression problems. In this paper we propose a bootstrap approach to
constructing critical points for use in outlier detection in the context of least-squares Studentized residuals, and find that this approach allows naturally for mild departures in model assumptions such as non-Normal error distributions. We illustrate our methodology through both a real data
example and simulated data.
Keywords: Case-based resampling; RSTUDENT; error distribution; externally Studentized residuals; internally Studentized residuals; jackknife-after-bootstrap; residual-based resampling
Document Type: Research Article
Affiliations: Australian National University, Canberra, Australia
Publication date: 01 August 2006
- Access Key
- Free content
- Partial Free content
- New content
- Open access content
- Partial Open access content
- Subscribed content
- Partial Subscribed content
- Free trial content