Hi,
I am carrying out my first linear regression project at university. I have political stability as my DV and quality of infrastructure as my IV. I am attempting to put together a list of control/confound variables to add to the model.
I am unsure though what exactly constitutes a control variable and am struggling as I believe some of mine may be mediators. So far I have chosen:
Freedom from corruption
Property rights
Policies toward foreign investment
Wastefulness of Gov spending
Rule of law
Annual population growth
Government debt.
I apologise if this is a silly question but any help would be much appreciated.
Many thanks,
Joe
A control variable is a variable you introduce in your model different from the DV or the key variable you want to check. The sign to be expected or unexpected sign just depend on their definitions. Of course, always have in mind the significance. Be careful, because yo could have some multicollinearity in your model.
Thank you for your reply. Please could you advise me further on the multicollinearity ? My understanding is that meas that two of my IVs are correlated - is that correct?
Two or more. You have clear candidates as you use 4 institutional quality indicators and other development variables. You can check for it. It is easy with any econometric software. For example, VIF or other tests. if so, you might eliminate some variables.
Thank you for your help, I am new to this. I ran VIF last night on the model and the highest figure was 8, for corruption. I researched a bit about the tolerances and found no clear agreement. Most said that up to 10 was fine, but some said 6 or even 4 was the limit.
You have several multicollinearity test. Try using Farrar & Glauber or CNT. So you check different options.
I think you use all institutional indicators as dV, no problems of multicolinearity with them if so. Good luck and relax. In econometrics you need to expend time. It is a meticulous work. Best whishesv
- Badges
- Science topic
- Similar topics
- Geoscience
- Asia
- India