Page 128 - Contributed Paper Session (CPS) - Volume 2

CPS1461 Michal P. et. al

Changepoints in dependent and heteroscedastic
panel data
2
1*
1
Michal Pešta , Matúš Maciak , Barbora Peštová
1 Charles University, Faculty of Mathematics and Physics, Department of Probability and
Mathematical Statistics, Czech Republic
2 The Czech Academy of Sciences, Institute of Computer Science, Department of Medical
Informatics and Biostatistics, Czech Republic

Abstract
Detection procedures for a change in means within a very general panel data
structure are proposed. Unlike classical inference tools used for the
changepoint problem in the panel data framework, we allow for mutually
dependent panels, unequal variances across the panels, and possibly an
extremely short follow up period. Two competitive self-normalized test
statistics are introduced and their asymptotic properties are derived for a large
number of available panels. The proposed tests are shown to be consistent
and their empirical properties are investigated.

Keywords
panel data; dependence; changepoint; heteroscedasticity; self-normalized test
statistics

1. Introduction
Panel data typically occurs in situations where some covariate of interest
is repeatedly measured over time simultaneously on multiple subjects—panels
(for instance, a financial development of a set of companies, economic growth
of some specific countries, or some qualitative performance of various
industrial businesses). For such data generating mechanisms, it is also
common that sudden changes can occur in the panels and especially the
common breaks in means are wide spread phenomena. These changes are
caused by some known or unknown causes and the statistical models used for
the panel data estimation should have the ability to detect and estimate these
structural breaks. Another crucial task is to decide whether the changepoints
are indeed present in the underlying panels, or not.

* Corresponding author (michal.pesta@mff.cuni.cz).
The research of Michal Pešta was supported by the Czech Science Foundation project GACˇ R
No. 18-01781Y. Institutional support to
Barbora Peštová was provided by RVO:67985807. Financial support through the Czech Science
Foundation project GACˇ R No. 18-00522Y
is gratefully acknowledged by Matúš Maciak.
117 | I S I W S C 2 0 1 9