Chow Breakpoint test

[Fang]

Hi I have a time series data from 1990 to 2012. After estimation, I want to test if there the result is stable so I used the Chow Breakpoint test. I picked up year 2002 as a breakpoint. Eviews provides three tests under this category: F-statistic, Log likelihood ratio, and Wald statistic. I found the estimation passed F-statistic, and failed the rest two tests. In this case, may I conclude that the estimation is stable because the F-statistic is good? Any comment? Thanks a lot.

[startz]

The difference is because you have a small sample. All three statistics will be about the same in a large sample, but not necessarily in a small sample. There really isn't a terribly good answer as to what to do when the three substantially disagree.

[Fang]

Thanks a lot for your input. Then in my case, what test should I use to test if there is any structure change? Thanks.