Multicolliniearity Problem

Moderators: statman, Analyst Techy, andris, Fierce, GerineL, Smash

ThesisWriter2012
Posts: 2
Joined: Tue Jul 24, 2012 6:18 am

Multicolliniearity Problem

Dear All,

my professor requires me to test for u-shaped relationships in order to graduate. Unfortunately my statistical education has no coverred regression models and everything I know so far, I have learned by readings blogs or books.

I have three different experience variables that are supposed to influence firm performance. Two of them are significantly correlated to the firm performance variable, one is not.

The one that is not significantly correlated, when I standadrize it with the function in SPSS (Descriptive, Save as standardized) or mean center it and then square it, in order to test for the u-shaped relationship. Then it has a very high VIF (ca. 28). After I remove the squared term of the other experience variable, it still remains with 10,4, which is a bit too high. The same problem occurs when I test for a moderating effect. Authors that have used it in the same context before managed to still find a u-shaped relationship although it has not been correlated to firm performance.

Oh and I need to lag my dependent variable by 1 in order to correct for autocorrelation. I do not know if that has an impact on the multicolliniearity problem.

Is there anything I can do to solve this problem?
I would appreciate any help Thanks a lot
Penguin_Knight
Posts: 473
Joined: Thu Apr 05, 2012 5:58 pm

Re: Multicolliniearity Problem

ThesisWriter2012 wrote:The one that is not significantly correlated, when I standadrize it with the function in SPSS (Descriptive, Save as standardized) or mean center it and then square it, in order to test for the u-shaped relationship. Then it has a very high VIF (ca. 28). After I remove the squared term of the other experience variable, it still remains with 10,4, which is a bit too high. The same problem occurs when I test for a moderating effect. Authors that have used it in the same context before managed to still find a u-shaped relationship although it has not been correlated to firm performance.
High VIF always happens to at least two predictors. You said that the VIF of the centered squared one is 28, what is the other independent variable(s) with high VIF? It'd help if you can list the variables and their VIF in each model.
ThesisWriter2012 wrote:Oh and I need to lag my dependent variable by 1 in order to correct for autocorrelation. I do not know if that has an impact on the multicolliniearity problem.
I am afraid no one would know until you look at it.

Who is online

Users browsing this forum: No registered users and 1 guest