The article compares different types of regression models: parametric, semiparametric, and nonparametric. Parametric models make strong assumptions but have well-known estimation methods like least squares. Nonparametric models don't make assumptions about relationships but are computationally expensive. Local linear smoothing is better than local constant for bias at data boundaries. Robust estimation methods for nonparametric models are limited. A new robust estimation method for semiparametric and nonparametric models is introduced.