Navigation

Recent Publications

More...

Home

Unit root testing on Buffered Autoregressive Model

diwang - Posted on 29 May 2017

Project Description: 

In this research, we aim to develop bootstrap-based unit root testing method for the newly proposed Buffered Auto-Regressive (BAR) model (Li 2016). In this research, a large number of simulations have to be implemented to verify the size and power of our proposed methods with different algorithms. In addition, several real data analysis in macroeconomics and finance will be implemented to provide new interpretation based on the new model.

Research Project Details
Project Duration: 
01/2017 to 12/2017
Project Significance: 
This research project will contribute and complement the research in nonlinear time series study. The previous study on BAR model are mainly based on the assumption that the time series data is stationary and our research project provides statistical tools to test whether the data can be viewed as stationary. The theoretical and empirical results that we have already got can show the great flexibility and potential of the BAR model. It would be interesting and important to show the performance of the proposed method based on numerical simulation
Results Achieved: 
The asymptotic properties of our proposed test statistics have been obtained. Further, we need to implement numerical simulations to verify the size and power of the proposed tests for the finite sample.
Remarks: 
Our proposed bootstrap-based method depends on a big amount of numerical simulation and complicated nonconvex optimization which require lots of computing resources. HKU HPC helps to provide computing resources and accelerate our simulation results.