0

I have a simple question. I'm doing a regression with countries (346 countries). I have a variable that measures level of previous conflict. I rescaled this variable in a variable that goes from 0.0 (minimum) to 1.0 (maximum) (mean 0.16 and SD = 0.23). I found that I need a quadratic and a cubic term of this variable. My dependent variable is attitudes towards abortion at the national level and it goes from 0 to 9. Now I found that my intercept is 1.74, my main term (conflict) is -7.00, my quadratic term is 17.32 and my cubic term is -10.15. How do I interpret this?

These are raw polynomials.

Jan Modus
  • 45
  • 9
  • Has conflict been analysed as a "raw" or an orthogonal polynomial? – Ian_Fin Aug 16 '16 at 10:47
  • It is a raw one. – Jan Modus Aug 16 '16 at 10:48
  • Nope, it isn't. Im not using orthogonal polynomials. This output comes from Mplus and not R. – Jan Modus Aug 16 '16 at 14:07
  • The statistical content remains the same whatever software you're using. If you need software help, please consult our help center for a list of support links. Is there a question that you have which is not answered [here](http://stats.stackexchange.com/questions/211534/how-to-interpret-quadratic-terms), with respect to quadratic regression? Because there are many more possible polynomial models than there are CV users, it would seem that the distinction between second- and third-degree polynomials is not as important as understanding how regression with basis expansion works generally. – Sycorax Aug 16 '16 at 14:42
  • Your explicit question is well answered in the linked thread. The problem, I suspect, is not w/ the question you actually asked. You should not be using linear regression w/ a response that is 0 to 9. You need to use ordinal logistic regression (& interpret the polynomial coefficients as in the linked thread), then you'll be fine. – gung - Reinstate Monica Aug 16 '16 at 16:08
  • Ok, I did that. I get +- same results. And no, my data is without mistakes or high / low values. In my view, it has to do with 0 to 1 coding of the variable. – Jan Modus Aug 16 '16 at 16:49

0 Answers0