Semiparametric methods in econometrics (Q1389388): Difference between revisions

Semiparametric methods are rather fresh tools in econometrics. Most of the relevant literature was published in the second half of the eighties and in the nineties. Semiparametric methods fill the gap between classical parametric and nonparametric methods. Standard methods for estimating empirical models in economics rely heavily on assumptions about functional forms and the distribution of the unobserved error process. Common are the assumptions of linear functional forms as well as normally distributed errors. As the author argues these assumptions are often violated in applied research or cannot derived from economic theory. In the first case the estimates may lead to wrong results and conclusions drawn from them. Therefore, the author suggests the application of semiparametric methods. Models and estimation problems that involve an unknown function and an unknown finite-dimensional parameter are called semiparametric. Most parts of the book are technical and the reader requires a sound background in mathematics, statistics as well as econometrics. As far as possible, examples are given to illustrate the working of semiparametric methods. Although semiparametrics are applicable to time series econometrics as well, most of the models and techniques are devoted to cross section or panel data. Special models presented in the book are the single-index models, binary response models, and transformation models. The book gives an excellent survey on a new field of research in econometrics and is recommended to all applied researchers who deal with cross section or panel data as an alternative approach to standard microeconometrics.