Change Point Analysis of Correlation Structure: Evidence from Selected Real Data and Monte Carlo Simulation
Author: Sharia Iqbal

The motivation of change point detection is to take into account sudden changes in distribution of time series, which remain important in social sciences, economics, medicine, environmental sciences, etc. In addition, due to distributional changes some statistical properties of time series changes such as variance, mean and correlation. The present study has been devised to explore the abrupt distributional changes in time series in multivariate frame work using non parametric method i.e. E-Divisive method, to detect multivariate change point. Our main focus is to explore the abrubt distributional changes in macro data of paskistan in multivariate framework. To check the performance of E-Divisive method for different location and number of change points in case of multivariate and individual analysis; we have applied Monto Carlo simulations. It has been found that the performance of E-Divisive method has high power in experiments locating change points at same and varying locations in multivariate scenario. In this study, attempt has been made to explore the change point in multivariate time series i.e. Macro-economic variables from energy sector and banking sector of Pakistan economy are selected. Results are drawn by assessing correlations among macro variables, unit root stationarity tests and change point detection method i.e. E-Divisive. The study spans over a period from 1990-2016. The study uses variables like production of Electricity, production of Natural gas, production Crude Oil, Cash in Pakistan, Balances with State Bank of Pakistan, Borrowing from State Bank of Pakistan, Real interest rate, Exchange rate and M2 (money supply). Change point have been identified in all of the series under study in univariate as well as multivariate framework. Supervised by: Dr. Amena Urooj Co-Supervised by: Ms. Uzma Zia

Meta Data

Keywords : Banking Sector, Change point detection, E-Divisive, Energy Sector
Supervisor: Amena Urooj
Cosupervisor: Uzma Zia

Related Thesis​