LR test for a structural break

[dan onath]

Hi,

I am using an ordered logit/probit model (ologit / oprobit) and would like to examine whether there is a structural break that affects two of my variables. My understanding is that a LR test is well-equipped to do this, but I am not sure how I should handle my other (control) variables re: the restricted and unrestricted models.

For example; my variables follow:

• y = regressand
• x1 x2 = variables of interest
• break = dummy variable that is 1 for dates after my expected structural break
• controls = several control variables

My main question is: if I am only looking for a structural break for x1 and x2, should I keep the interaction term between break* and my control variables? E.g., should I do this...

ologit y break##(x1 x2 controls)
estimates store unrestricted
ologit y x1 x2 break#(controls)
estimates store restricted
lrtest unrestricted restricted

or just...

ologit y break##(x1 x2 controls)
estimates store unrestricted
ologit y x1 x2 controls
estimates store restricted
lrtest unrestricted restricted

Please also let me know if the use of a LR test in this context is misguided.

Thanks.