Series estimation under cross-sectional dependence
Series estimation under cross-sectional dependence
复制标题
DOI：
10.1016/j.jeconom.2015.08.001
复制
发表时间：
2013
影响因子：
6.3
通讯作者：
Jungyoon Lee;P. Robinson
中科院分区：
文献类型：
--
作者：
Jungyoon Lee;P. Robinson
在允许截面相关性和异质性(包括条件和无条件异方差)的扰动条件下，以及允许相关性和不需要存在密度的回归变量的条件下，发展了横截面数据非参数和半参数回归模型的级数估计的渐近理论。这些条件旨在适应经济应用中看似合理的各种设置，也适用于面板、空间和时间序列数据。建立了一个非参数回归估计的均方收敛速度，然后是一个相当一般的统计量的渐近正态分布，证明了依赖于单指标或双指标来对数据进行排序的数据驱动的学院化是合理的。在部分线性模型的设置下，进行了有限样本性质的蒙特卡罗研究和两个经验应用。
An asymptotic theory is developed for series estimation of nonparametric and semiparametric regression models for cross-sectional data under conditions on disturbances that allow for forms of cross-sectional dependence and heterogeneity, including conditional and unconditional heteroscedasticity, along with conditions on regressors that allow dependence and do not require existence of a density. The conditions aim to accommodate various settings plausible in economic applications, and can apply also to panel, spatial and time series data. A mean square rate of convergence of nonparametric regression estimates is established followed by asymptotic normality of a quite general statistic. Data-driven studentizations that rely on single or double indices to order the data are justified. In a partially linear model setting, Monte Carlo investigation of finite sample properties and two empirical applications are carried out.