Robust Statistics

Read these first

1. 01Huber's Contamination Model

   The formal statement of 'some small, unknown fraction of my data might be garbage of an unknown kind', the assumption that everything in robust statistics is built to survive.

2. 02Outlier-Resistant Mean and Covariance Inputs

   Practical estimators of a mean vector and covariance matrix that keep a handful of contaminated observations from silently corrupting every downstream portfolio and risk number.

Then the rest

1. 03Additive Versus Innovation Outliers in Time Series
2. 04Gaussian Rank Correlation for Robust Covariance
3. 05Iteratively Reweighted Least Squares
4. 06Least Trimmed Squares and Least Median of Squares
5. 07MM-Estimators for Regression
6. 08S-Estimators for High-Breakdown Regression
7. 09The Median-of-Means Estimator
8. 10The Robustness Versus Efficiency Tradeoff
9. 11The Sandwich Variance Estimator for M-Estimators
10. 12Tukey's Biweight and Redescending Psi Functions