Nonparametric and Semiparametric Estimation in Complex Surveys

摘要: Publisher Summary This chapter focuses on nonparametric and semi-parametric methods in two important statistical areas: estimation of densities regression functions. Both these areas have applications survey estimation, for both descriptive analytical uses. Orthogonal decomposition is a non-parametric method with good properties that applicable situations where the mean function not necessarily smooth. Neural networks are class conceptually related to penalized spline regression, which parameters found by nonlinear regression. The model particularly useful when some covariates data set categorical, definition cannot be smoothed. In addition multivariate data, another extension models more complex structures, including equivalents generalized linear models. Nonparametric require specification one or several smoothing such as bandwidth kernel penalty