Sometimes one wants to model the effect of a variable as a nonlinear smooth curve. A convenient choice for such a curve is a restricted cubic spline. This option has existed in Stata for a while through user-written programs, but as of Stata 10, the mkspline command in combination with the cubic option has been implemented in official Stata. In this talk, I will briefly introduce splines and restricted cubic splines, but I will mainly focus on what happens after one has estimated a model with a restricted cubic spline, and in particular how the postrcspline package can help in the interpretation of the results.
