Projection-based outlier detection in functional data

We propose a procedure based on a high-breakdown mean function estimator to detect outliers in functional data. The robust estimator is obtained from a clean subset of observations, excluding potential outliers, by minimizing the least-trimmed-squares projection coefficients after functional principal component analysis. A threshold rule based on the asymptotic distribution of the functional score-based distance robustly controls the false positive rate and detects outliers effectively. Further improvement in power can be achieved by adding a one-step reweighting procedure. The finite-sample performance of our method demonstrates satisfactory false positive and false negative rates compared with existing outlier detection methods for functional data.