摘要
空间计量经济学发展了一系列的空间计量模型,但在实证研究中如何根据实际问题选用最佳模型,尚无统一的筛选框架。而且既有文献中用于模型选择的检验主要针对大样本情形,但现实并不总是满足大样本的前提条件。文章结合贝叶斯理论处理有限样本的优势构建了针对空间计量经济模型进行筛选的统一框架。空间计量模型Bayes筛选框架的突出优势是在处理嵌套模型与非嵌套模型两种情形时具有逻辑一致性,对大、小样本条件均适用,而且筛选指标具有可直接计算性,易于操作。算例计算结果表明:相对其他所给模型而言,整合了网络技术联系及经济联系的空间误差自回归模型更适于中国电信服务外溢性的研究,这与传统空间计量经济方法分析结果一致。
A whole family of different spatial econometrics model has been developed, but how to choose the ＂best＂ model using appropriate criteria to meet the need of empirical research. There is no uniform and coherent framework for such model choice. Furthermore, existing test for model choice aim at large sample-sizes mostly, which result in the biased estimation from finite-sample. This paper develops a comprehensive and coherent Bayesian theory framework for spatial econometrics model choice, including both nested, non-nested, finite-sample and large-sample sizes models within the choice set. Finally, this framework applied to Spatial Econometrics model choice on spillovers of Chinese Telecom services. Findings indicate that spatial econometrics model choice framework based on Bayesian theory is coherent in nest model and non-nest model; there are robust performances both in finite -sample and large-sample sizes. Moreover choice criterion can be computed directly, do not need simulation. Research discovered that the spatial autoregressive error model integrated network neighbor effect and economic relation is more suit the researches on spillovers of Chinese telecom services than the others. This consistent in classical methods of spatial econometrics.
出处
《华东经济管理》
CSSCI
2008年第10期61-64,共4页
East China Economic Management
