Nonparametric Regression Analysis of Growth Curves

摘要: In recent years, nonparametric curve estimates have been extensively explored in theoretical work. There has, however, a certain lack of convincing applications, particular involving comparisons with parametric techniques. The present investigation deals the analysis human height growth, where longitudinal measurements were collected for sample boys and girls. Evidence is presented that kernel acceleration velocity height, itself, might offer advantages over fitting via functional models recently introduced. For specific problem considered, both approaches are biased, but one shows qualitative quantitative distortion which not easily predictable. Data-analytic problems involved estimation concern choice kernels, smoothing parameter, also whether parameter should be chosen uniformly all subjects or individually.