Assume I have a set of data and would like o apply multiple linear regression. As we know multiple linear regression has assumptions such independent variables shouldn't be highly correlated (coefficient <0.8). My question is what to do when assumptions dont hold? Apply nonlinear regressin? If so, how to decide which form to apply?

  • 1
    $\begingroup$ Highly correlated variables would be probably among my lowest concern. I believe you just get inflated standard errors. Of course independent variables are going to be correlated. This is an overplayed concern. $\endgroup$
    – John Stud
    Feb 4 at 17:30