Test by adaptive Lasso quantile method for real-time detection of a change-point (Q1669885)

The author is concerned with the problem of detecting changes in the parameter vector of a linear model in real-time. In this, she considers the case of a non-standard error distribution as well as the presence of a large number of regressors. The proposed methodology relies on adaptive least absolute shrinkage and selection operator (Lasso) quantile estimation. For this estimation, historical data are used which are assumed to be generated by a common model (without a change-point). For the subsequent sequential change-point detection, a test statistic is defined which is based on the residuals under the fitted model. The author works out the asymptotic null distribution of the proposed test statistic, and analyzes consistency properties of the resulting change-point test under alternatives. Finally, two applications to real-life data are presented.