2Q Introduction to Time Series Regression Analysis
Marco Ercolani, University of Birmingham
23 July - 3 August (two week course / 35 hrs)
Detailed Course Outline [PDF]
Course Content
This is a course for graduate students, academics and researchers and who want an introduction to time-series regression analysis. In the course we seek a balance in our understanding between the theory and the application of time-series regression analysis. We mostly use the statistical software package GRETL because it includes most of the statistical procedures we need, because it is freely available, and because it can read data from EViews, R, SPSS and Stata.
We start the course with a basic introduction on the big topics in time-series regression and why different researchers have different approaches. We also see how best to present our time-series data in graphs that can be included in our research work. We then move on to basic tests (e.g. Durbin-Watson) to detect if there is autocorrelation in the errors of the model we have estimated. We will learn to model this autocorrelation either by including time-lags of variables or by estimating models where the error terms are autoregressive (e.g. by Cochrane-Orcutt estimation). We then proceed to slightly more advanced techniques such as Vector Auto-Regression (VAR) estimation and its application to Granger tests of causality between variables. We also see how to carry out tests for structural stability of the model (e.g. the Chow test). Finally we move on to tests of variable stationarity (e.g. Dickey-Fuller tests) and to basic tests of cointegration between non-stationary variables (e.g. Engle-Granger test).
The course comes with a booklet that provides the script commands needed to carry out all these tests in Gretl, EViews, R and Stata. This booklet is also useful as a reference book after the course has ended. Datasets used for the course include data on global warming, macroeconomic consumption, voting intention in the 2010 UK election, wind-turbine generation and on price inflation. There is also scope for participants to analyse their own datasets should they wish.
Course Objectives
Participants should become familiar with popular techniques commonly used in time-series regression analysis and how to apply them. With this knowledge they should be able to choose the statistical technique most appropriate to the particular research they are carrying out. They should also be able to interpret a wide range of statistical procedures that are often encountered when reading research papers.
Course Prerequisites
Participants need to be familiar with basic (non time-series) regression techniques. This includes understanding OLS regression, R2 statistics, t-statistics and F-statistics.
Background Reading
James Stock and Mark Watson 2011 “Introduction to Econometrics”, 3rd edition.
Gary Koop 2008 “Introduction to Econometrics”
Dimitrios Asteriou and Stephen G. Hall 2011 “Applied Econometrics” 2nd edition
Required Reading
(No purchase necessary before the course, all the books are available in the Summer School Library)
Box, G.E.P., and Jenkins, G.M. 1994. “Time series analysis: forecasting and control”, 3rd ed.
Chatfield, C. 1996. “The Analysis of Time Series: An Introduction”, 5th ed.
Gourieroux, C., and Monfort, A. 1997. “Time Series and Dynamic Models”.
Hamilton J. 1994. “Time Series Analysis”.
Maddala, G. 2001. “Introduction to Econometrics”, 3rd ed.
Ostrom, C. 1990. “Time Series Analysis: Regression Techniques”, 2nd ed.