DETECTION OF STRUCTURAL CHANGE IN ECONOMICAL TIME SERIES
Time series are frequently affected by different unexpected events such as revolution, war, strike, etc. We call these external events interventions (or structural change) which should be detected and considered carefully in statistical analysis.In this research both parametric and non-parametric procedures are considered to detect and localize these change points. The first one is based on a classical approaches, i.e. the Recursive residual, Cusum of square, QuantÂs log likelihood and Likelihood Ratio Test (LRT). However, in many practical situations, the explicit form of the probability density function is difficult to determine and an assumption of a specific distribution may yield misleading results. To overcome this problem a non-parametric procedure based on the norm of the difference between two spectral density functions is derived. Finally, the demand for money over 1959-1997 period in Iran are employed to demonstrate the applicability of the detection procedures.
Year of publication: |
2000-07-05
|
---|---|
Authors: | Yarmohammadi, M. |
Institutions: | Society for Computational Economics - SCE |
Saved in:
Saved in favorites
Similar items by person
-
Monotonic Power in tests for structural change in the mean based on orthonormal series filtering
Andreou, Elena, (2006)
-
E-consumers' search and emerging structure of B-to-C coalitions
Laye, Jacques, (2006)
-
Macroeconomic factors in the term structure of interest rates when agents learn
Laubach, Thomas, (2006)
- More ...