Don’t Let Messy Data Bend Your Curves

This content explains how to accurately fit curves to data that contains outliers or other messy observations. It contrasts classical methods like Ordinary Least Squares (OLS), which are easily distorted by spurious data, with robust estimation techniques that resist their influence. Two primary methods for modeling non-linear relationships are detailed: polynomial regression and splines. While polynomial regression uses higher-order terms to create flexible curves, its global nature can be a drawback, whereas splines offer a more localized approach by joining piecewise polynomial functions at specific points called knots.