Quandt-Andrews Test Error

[kevinliberty]

FYI I am receiving the same data error for the sample date range in the Chow Breakpoint Test.

[kevinliberty]

They both have HAC errors.

[EViews Gareth]

EViews 6 had a bug in it that meant it didn't calculate the Quandt Andrews test with HAC errors correctly. EViews 7 does.

[kevinliberty]

As a hypothetical, if I was running EViews 6 and I was using the probability of a the Quandt-Andrews Breakpoint Test w/ HAC errors as the determining factor of my sample data range then the probability that was being reported in EViews 6 would not be accurate?

So in my previous example the QA Breakpoint Test determined the the probability of the largest value was 13.44626 with a Prob of .13333. That probability value is inaccurate?

[EViews Gareth]

Right.

[kevinliberty]

Okay thanks,

Can you let me know when the Chow Breakpoint Test bug is fixed as well?

[EViews Gareth]

With the EViews 8 release, EViews 7 patches are reducing in frequency. Probably won't be for a while.

[kevinliberty]

Is there a way to retrieve the df1, and df2 for the QA breakpoint test?

[EViews Gareth]

What do you mean by DF1 and DF2?

[kevinliberty]

Normally with an F-stat test there are 2 different degrees of freedom associated with the F-stat test value and the resulting probability.

[EViews Gareth]

Correct, but the Quandt-Andrews test performs many F-tests.

[kevinliberty]

Okay so is there a way to retrieve the degrees of freedom for the specific date that is referenced in the QA test results?

[EViews Gareth]

No, they aren't calculated/used.

[kevinliberty]

How is the p-value found for the F-stat value found then?

[EViews Gareth]

As mentioned on the output:

Code: Select all

Note: probabilities calculated using Hansen's (1997) method

The User Guide has more details.