Are tests for smooth structural change affected by data inaccuracies?
Author
Summary, in English
The size and power of tests for smooth structural change are evaluated in the presence of random measurement error in the explanatory variable or outliers in the dependent variable of a univariate regression model. It is shown that the considered tests are robust to measurement error of a magnitude that can be found in real economic data. By contrast,outliers are found to distort both the size and the power of test for structural breaks. It is shown that the effects of outliers can be compensated by a simple wavelet-based outlier detection algorithm.
Department/s
Publishing year
2012
Language
English
Full text
- Available as PDF - 304 kB
- Download statistics
Document type
Student publication for Master's degree (one year)
Topic
- Business and Economics
Keywords
- Structural breaks
- measurement error
- additive outliers
- wavelet analysis
Supervisor
- Yushu Li (Dr.)
- Fredrik N G Andersson