Discussion of “The power of monitoring: how to make the most of a contaminated multivariate sample”
- 73 Downloads
In this discussion, we consider two examples. The first example concerns the Old Faithful data, which the authors (Cerioli, Riani, Atkinson, Corbellini in Stat Methods Appl, to appear) discuss in detail in their paper. The second example, which is taken from www.kaggle.com, is based on the prices and other attributes of 53,900 diamonds. The point of our discussion is to demonstrate that the process of producing valid models and then looking at diagnostics, that compare least squares and robust fits, can also effectively identify outliers and/or important structure missing from the model. Using this approach, we identify a dramatic change point in the diamonds data. We are very curious about what information the sophisticated monitoring methods produce about this change point and its effects on the outcome variable.
KeywordsHigh breakdown Rank-based Robust Robust diagnostics Wilcoxon
We would like to thank an anonymous referee for his suggestions on clarification of several passages of the paper.
- Cerioli A, Riani M, Atkinson AC, Corbellini A The power of monitoring: How to make the most of a contaminated multivariate sample. Stat Methods Appl (to appear)Google Scholar
- McKean JW, Hettmansperger T (2016) Rank-based analysis of linear models and beyond: a review. In: Liu R, McKean JW (eds) Robust rank-based and nonparametric methods. Springer, Berlin, pp 1–24Google Scholar