1
$\begingroup$

I have a set of data, sampled once per year from 1990 to 2008. It looks more or less linear. However, doing a Chow test results in a significant p-value (0.003 and below), wherever I test for a structural break. What is going on?

(There was a similar question once before, but it didn't really receive a satisfactory answer.)

Here is a screenshot showing what I am doing.

The p-value is very small wherever I test to see if there is a structural break.

An answer would be much appreciated!

$\endgroup$
1
$\begingroup$

If the "true" relationship is a curve then two straight lines will fit better than one straight line, wherever the break.

This is the basis for what Minitab calls a "data subsetting lack of fit test". Although Google found a statement that "It’s hard to find a description of precisely what Minitab does for it’s data subsetting", the Minitab Methods and Formulas Help is explicit, showing all the calculations and referring to D.C. Montgomery and E.A. Peck (1982). Introduction to Linear Regression Analysis. John Wiley & Sons.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.