Re: Bai and Perron Sequential test
Posted: Wed Mar 02, 2022 10:42 am
by EViews Gareth
Again, without wanting to go into deep mathematical issues (lookup something like QR updating if you're interested), once you have computed the RSS for the overall regression, or a single subset regression, if you want to know the RSS that would result from adding or removing an observation (which is what you're doing when testing different break dates), you do not need to calculate the coefficients themselves, you can just "update" the RSS based upon the previous one.
Re: Bai and Perron Sequential test
Posted: Wed Mar 02, 2022 10:43 am
by EViews Gareth
Think of it this way.
If I ask you to compute 3402*151 without a calculator, and then I ask you to compute 3402*152 without a calculator - when doing the second calculation, would you start from scratch, or would you use the result from the first and update it by adding another 3402?
Re: Bai and Perron Sequential test
Posted: Fri Apr 28, 2023 12:09 pm
by EViews Glenn
Just noticed this discussion. As Gareth says, the coefficients are not computed as this would make the computation prohibitively slow. There is a large litererature on this topic. This may help clarify.
https://onlinelibrary.wiley.com/doi/ful ... 02/jae.659
and
Brown RL, Durbin J, Evans JM. 1975 Techniques for testing the constancy of regression relationships over time. Journal of the Royal Statistical Society, Series B 37 149–192.