The article discusses methods for estimating densities and regression functions in surveys. Nonparametric regression, like orthogonal decomposition, is useful when the mean function is not smooth. Neural networks, similar to penalized spline regression, are used for finding parameters through nonlinear regression. Semi-parametric models are handy when some data covariates are categorical. Nonparametric regression can be extended to models with complex mean structures, like generalized linear models. Smoothing parameters, such as bandwidth in kernel regression, are crucial for nonparametric regression applications.