Estimation and bootstrapping under spatiotemporal models with unobserved heterogeneity
提出一种新的时空模型，通过分位数函数相关随机效应和异方差创新刻画个体未观测异质性，采用两阶段估计和混合双重自助法，适用于城市空气质量等空间数据分析。
Proposed herein is a novel spatiotemporal model to characterize the unobserved heterogeneity across individuals using quantile-function-based correlated random effects and heteroscedastic innovations in a general framework. This model can be used to explore the influence of space-specific factors on latent effects at different quantile levels by controlling for spatiotemporal effects. A two-stage estimation procedure is introduced in which (i) the method of moments is used to estimate spatiotemporal effects then (ii) quantile regression is used for individual effects. A hybrid double bootstrapping procedure is then proposed to approximate the asymptotic distributions of coefficient estimators. The validity of the estimation and bootstrapping is established theoretically and then confirmed by simulation studies, and the usefulness of the proposed model is demonstrated with a real example involving city air quality.