Help on a project!!!

I am doing a school project and trying to find a relationship between government deficit/GDP (dependent variable) and inflation, interest rate, employment rate(three independent variable). After I did some test in gretl, the linear regression model passed REMSEY test for non-linearity, but failed to pass the CHOW test and White test. My question is what my next step should be. Adding/deleting a variable? dummy variable? log/first differencing variable? or look at ARIMA model? Any suggestions are greatly appreciated!
If non-linearity was detected, perhaps you should explore the transformations that would make the relationship approximately linear? You can transform the outcome or the predictor or both.