Semiparametric Estimation Using Pairwise Comparisons and Other 'Trimming' Methods

这个项目的贡献来自于为以前没有考虑过的模型开发半参数估计器，以及为统计和计量经济学文献中考虑过的模型开发新的估计器。计量经济学中的大量研究都致力于半参数方法的发展，因为这些方法允许分析师从数据中获得结果，即使对产生数据的潜在过程知之甚少。这个项目为许多不同的模型开发了半参数估计器。这些模型包括部分线性的Tobit模型、部分线性的Logit模型、部分线性的泊松回归模型、部分线性的久期模型和具有样本选择的Tobit模型。这些估计器是通过最小化基于比较所有观测对的目标函数来定义的。将对以这种方式定义的估计量的渐近性质进行调查。之前关于固定效应的Tobit模型的研究将继续下去。将讨论估计所谓类型3托比特模型的固定效果版本的可能性，并开发了一个两步估计器。将开发软件，允许其他人使用这些技术。

The contribution of this project comes from the development of semiparametric estimators for models that have not previously been considered, as well as new estimators for models that have been considered in the statistical and econometric literature. A great deal of research in econometrics has been devoted to the development of semiparametric methods because these methods permit analysts to obtain results from data even if very little is known about the underlying processes that generated the data. This project develops semi-parametric estimators for a number of different models. These include partially linear Tobit models, partially linear logit models, partially linear Poisson regression models, partially linear duration models, and Tobit models with sample selection. These estimators are defined by minimizing objective functions that are based on comparing all pairs of observations. An investigation of the asymptotic properties of estimators that are defined in this way will be conducted. Previous research on Tobit models with fixed effects will continue. The possibility of estimating fixed effects versions of the so- called Type 3 Tobit model will be addressed, and a two-step estimator is developed. Software will be developed to permit others to use these techniques.