Department of Statistics Seminar
North Carolina State University

presents

Dr. David A. Dickey

North Carolina State University

"Stationarity Testing in Time Series"

ABSTRACT

In their 1970 text on time series, Box and Jenkins popularized the use of differencing data to make it conform to the standard assumptions (stationarity) for time series modeling as it was then practiced. For example, one might model the differences of the Dow Jones Industrial Average rather than the levels. Both numbers are reported on nightly news programs. This differencing operation implies a unit root in the characteristic polynomial of the levels, a testable hypothesis. The difficulty with testing for unit roots is that nonstandard distributions are involved. Many economic hypotheses lead to unit root tests and once these nonstandard distributions were tabulated, there began a flood of economics papers applying unit root tests. Interest in unit root problems spans a couple of decades, with sessions on unit roots often appearing in the programs of national meetings. I will give an introductory overview at a level accessible to graduate students rather than a detailed report on our most recent work. I will point out the statistical problems involved, give several examples, talk about the relationship to cointegration, and briefly review some relatively recent extensions.

Friday, September 15, 2000

3:35 - 4:35 pm

206 Cox Hall

Refreshments will be served on the second floor of Dabney Hall (left of Room 222) at 3:00 pm.