If I get two break-dates from Bai-Perron structural break test but the history supports only one, then should I follow the test? Or, check for another test?
It depends on the situation. If you know that there is only one break, you would presumably know when the break occurs. So why use a test for unknown break points? If, on the other hand, you are testing to see if the historical interpretation matches the data, perhaps you could proceed by looking for another event that could trigger the break.
Also, it is worth considering that the breaks you are finding may be an indication of model mis-specification. Could there be an important variable you are omitting or (more probably) is there nonlinearity in the relationship that your model does not account for?
Hi Monir,
Well, if the non-technical evidence clearly shows one break during the study period then it is worth following it rather than B-P test results. In this circumstance, you dont need to run the B-P test to identify the unknown breaks in the data series. My suggestion for you is that in your paper or project or dissertation provide a clear argument around the known break point in the data series and you may divide your sample data into two sub-groups (which should account before and after break) and then run the analyses on full sample period, before the break period and after the break period and highlight the results by comparing before and after break to see if there are any significant changes in the results. I believe my suggestions will help you to proceed further on ur project. All the best. Sudharshan
Hello Monir
You can find most important date of structural break in series by using Zivot-Andrews method. See page 733 of attached file for reasons and studies in which structural break are based on this method.
Regards
If Bai-Perron Structural break test suggests two break dates but the history supports only one, then you should follow the one that is supported by the history. The Bai-Perron test is a statistical test that can help identify structural breaks in time series data. However, it is important to note that statistical tests are not always perfect and can sometimes produce false positives or false negatives.
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Regression