Change-point estimators with the weighted objective function when estimating breaks one at a time

This study investigates the change-point estimators based on the recently proposed weighted objective function, wherein the model has two structural breaks and these breaks are estimated one at a time. The asymptotic distributions of the first-step and second-step estimators are investigated under the long-span asymptotic scheme, which is valid when the break sizes are large. We show that they are unimodal and asymmetric in general. However, for the small sizes of the breaks, the limiting distribution based on the long-span scheme cannot approximate the finite sample distribution; the finite sample distribution of the first-step estimator tends to have two peaks, while the corresponding asymptotic distribution under the long-span asmptotic scheme is unimodel. This finite sample property is approximated under the in-fill asymptotic scheme.

Graduate School of Economics, Hitotsubashi University

2023-11-14

English(eng)

technical report

VoR

59

Kurozumi's research was supported by JSPS KAKENHI Grant Number 22K01422 and 23H00804.

Discussion papers / Graduate School of Economics, Hitotsubashi University

https://hdl.handle.net/10086/80949