Significance of variables

[Joyce Parreira]

Hello,

I have one doubt about the significance of an independent variables. In my database I have economic related variables, political variables and oil price - all independents .

Some of my political variables in my regression are insignificant, however I was suggest to interact these insignificant variables with oil price. Is it correct? What I have in my mind is there is no point in doing that, since my "base" variable (the political) is insignificant. Is there any point in doing that?

Thanks in advance!

[Joseph Coveney]

You should set up your regression model to answer your research question. If your research question doesn't involve an interaction of oil price and political variables, then you're done.

Nevertheless, whowever it was who suggested forming interaction terms might have been trying to point out something like the following regarding your base variable's being "insignificant".

version 15.1

clear *
set seed `=strreverse("1475983")'

quietly set obs 20
generate byte oil = mod(_n, 2)
generate double pol = runiform()
generate double eco = 0 * pol + 0 * oil + (pol - 0.5) * oil + rnormal() / 5

regress eco c.pol i.oil
regress eco c.pol##i.oil

exit