Truncating CUSUM estimation for mean change-point in heavy-tailed dependent observation of panel data

Abstract

The problem of a truncating CUSUM estimation for mean change-point in heavy-tailed dependent observation for panel data is considered. The original sequence is truncated, and the variance after the truncation is limited; In the truncation case, a generalized Hájek-Rényi type inequality is derived and the truncating CUSUM estimation of the mean change-point is constructed. The consistency of the estimated change-point and the rate of convergence are established. The results show that the smaller the parameters of the heavy-tailed dependent sequence, the faster the convergence speed of the estimator. In order to have better robustness, it is a good approach for the truncating CUSUM estimation, which can weaken the influence of outliers.

Keywords

• heavy-tail dependent observations
• panel data
• truncating estimation
• mean change-point
• hájek-rényi type inequality