CMStatistics 2015: Start Registration
View Submission - CMStatistics
B1749
Topic: Title: Efficient estimation and testing of varying coefficient panel data models Authors:  Alexandra Soberon - Universidad de Cantabria (Spain) [presenting]
Winfried Stute - University of Giessen (Germany)
Juan Manuel Rodriguez-Poo - Universidad de Cantabria (Spain)
Abstract: Efficient nonparametric estimation and inference of varying coefficient panel data models is addressed. Our aim is twofold. On one hand, we propose a more accurate nonparametric estimator based on a pairwise differencing transformation. Later, a generalized least squares (GLS) procedure to incorporate the information of the error covariance matrix is presented. Despite the established theory in the literature, it is shown that the resulting feasible GLS estimator is asymptotically more efficient than the above without imposing any closeness property within groups. On the other hand, to estimate the variance coponents we develop a new technique based on the efficient estimation of moments when distributional assumptions such as normality of the error variables cannot be justified. Sufficient conditions for the asymptotic normality of these estimators are given. Furthermore, an alternative test statistic is proposed to test the significance of the individual effects based on the asymptotic properties of the estimator of the variance components. Some simulations are used to examine the finite sample performance of the proposed test statistic and estimators. Also, to see how useful are our results, we illustrate an application about the production efficiency in the European Union's companies.