Non linear regression and under-/overvaluation calc.

As you know the Security Market Line is used to find under- or overvalued securities (if they plot below or above the SML). I am doing something similar with a non linear regression that contains ln. Does that in any kind violate the ability to apply the over- or undervaluation logic to this regression? Sorry in case that is an odd question. I have got a humanities background only. Thanks guys!