A0713
Title: Uniform inference for nonparametric panel model with fixed effects
Authors: Nan Liu - Xiamen University (China) [presenting]
Abstract: The uniform inference is studied on a structural function $g(.)$ and its functionals in a nonparametric panel data model with individual fixed effects. This nonparametric model relaxes restrictions of time series behaviours by allowing for stationary mixingale or unit root regressors. After removing the individual fixed effects via within-group transformations, a sieve estimation method is proposed, with a sup-norm convergence rate established accordingly. Then, Yurinskii's coupling principle of Gaussian processes and the uniform confidence bands constructed by the sieve score bootstrap method are used to test for linear functionals of $g(.)$. Under the asymptotic framework of an increasing cross-sectional dimension and either a fixed or diverging time dimension, it is proved that the proposed Kolmogorov-Smirnov (sup-type) test has asymptotic uniform size controls and is uniformly consistent. Monte Carlo simulations confirm that the sieve estimator and its uniform confidence bands work well in finite samples. The above uniform inference procedure is also applied in one empirical setting, and some interesting results in nonlinear patterns of elasticity of consumption concerning income shocks are found.
