CMStatistics 2015: Start Registration
View Submission - CMStatistics
B0566
Topic: Contributions in time series and time-varying coefficients Title: Robust lasso regression using the Tukey's biweight criterion Authors:  Le Chang - the Australian National University (Australia) [presenting]
Steven Roberts - the Australian National University (Australia)
Alan Welsh - the Australian National University (Australia)
Abstract: The adaptive lasso is a common means of performing simultaneous parameter estimation and variable selection. The adaptive weights used in its penalty term mean that the adaptive lasso achieves the desirable oracle property. We propose an extension of the adaptive lasso called the Tukey-lasso. By using Tukey's biweight criterion, instead of squared loss, the Tukey-lasso is resistant to outliers in both the response and covariates. Importantly, like the adaptive lasso, the Tukey-lasso also enjoys the oracle property. The Tukey-lasso is compared to various competitors and extensive simulation studies show that it offers significant improvements in performance compared to the traditional adaptive lasso and other robust implementations of the lasso in the presence of outliers. Real data examples further demonstrate the utility of the Tukey-lasso.