Open Collections

A comparison of some robust estimates of correlations in the presence of asymmetry

University of British Columbia

Three methods for obtaining robust estimates of correlation matrices were compared in conditions of asymmetrically contaminated normal distributions. The estimators examined included a multivariate trimming (MVT) procedure based on the Mahalanobis distance measure, and two procedures derived from the robust estimation of regression coefficients. The regression coefficients were obtained from the sample cumulant generating function of the residuals. Monte Carlo results were obtained for various levels of sample size and outlier contamination. Correlations obtained from the regression procedures were observed to be highly robust with respect to asymmetric contamination and were able to withstand larger amounts of outlier contamination than the MVT estimates. The MVT estimates tended to be slightly less biased than correlations obtained from the regression procedure in conditions with the smallest amounts of contamination. The use of these estimates for outlier identification is discussed.

Text

2010-06-12

For non-commercial purposes only, such as research, private study and education. Additional conditions apply, see Terms of Use https://open.library.ubc.ca/terms_of_use.

http://hdl.handle.net/2429/25602

Psychology

University of British Columbia

Graduate