Different estimated coefficients when modeling the dummies

[sirooskhadem]

Hi,
I would like to show the behavior of a variable during specific hours of a day.

For example, for AUD we have a dummy variable (D_LOC) which covers from 23:00-06:00 and another dummy (D_FOR) which covers the remaining hours of a day, from 06:00-23:00.
Now problem is, I get different results when I model them together vs. when I model them separately. I have attached the outcomes.
Interestingly, when I model them separately the constant coefficients are same as the dummy coefficients when I model them together.

My question is which one should I pick?

Regards,
Siroos

[Carlo Lazzaro]

Siroos:
there's something that I don't understand about your query:
If the same dummy is used to split time in 6-23 and 24-5 periods, you should have something like:

dummy=0 if time ranges from 6 to 23;
dummy=1 otherwise.

But if you plug in among predictors two dummies (each representing one of the two time periods) it is strange that EViews does not drop one of them due to collinearity.

IMHO, you should better providing the list with some more details about what you're after.

Kind regards,
Carlo

[sirooskhadem]

Thank you Carlo, but I already dropped the constant, when use them together, so there should be no problem.
Cheers,
Siroos