Hello,
I am exploring the use of Digital Signal Processing methods in business time series analysis. I took up Hodrick-Prescott (HP) filter as my first case for no particular reason. I come across different reviews about the use of HP filter some favoring its use and some not favoring.
I understand that HP filter suffers from
- Inability to handle out liers, abrupt changes, and non-normal noise
- Arbitrary choice of the smoothing parameter λ
- Need for future values in forecasting
- Unstable behavior at the end of the series and hence ARIMA may be used to improve end-point estimation
- Introduces spurious cycles
I want to understand why HP filter is popularly used in economic time series analysis in spite of the above drawbacks? Are there any strengths that makes a compelling case? I will appreciate any guidance in this regard.
Thanks,
Vasu
------------------------------
Vasudevan Raghavan
Software Technology Analyst
------------------------------