| 8,135 | 243 | 180 |
| 下载次数 | 被引频次 | 阅读次数 |
空间计量模型的选择是空间计量建模的一个重要组成部分,也是空间计量模型实证分析的关键步骤。本文对空间计量模型选择中的Moran指数检验、LM检验、似然函数、三大信息准则、贝叶斯后验概率、马尔可夫链蒙特卡罗方法做了详细的理论分析。在此基础上,通过Matlab编程进行模拟分析。结果表明:在扩充的空间计量模型族中进行模型选择时,基于OLS残差的Moran指数与LM检验均存在较大的局限性,对数似然值最大原则缺少区分度,LM检验只针对SEM和SAR模型的区分有效,信息准则对大多数模型有效,但是也会出现误选。而当给出恰当的M-H算法时,充分利用了似然函数和先验信息的MCMC方法,具有更高的检验效度,特别是在较大的样本条件下得到了完全准确的判断,且对不同阶空间邻接矩阵的空间计量模型的选择也非常有效。
Abstract：Spatial econometric model selection is an important part of the spatial econometric modeling,is also a key step in spatial econometric model empirical analysis. We have made a detailed theoretical analysis on Moran index test,LM test,likelihood function,three information criteria,Bayesian posterior probability,Markov chain Monte Carlo method of spatial econometric model selection analysis. On this basis,we make simulation analysis by using Matlab programming.The results show that,in the model clusters of extended spatial econometrics,Moran index and LM test based on OLS residuals are some limitations,the principle of maximum log likelihood values is lack of differentiation,LM test is effective to only distinguish between SEM and SAR model. The information criterion is effective on most models,but also appear wrong choice. When M-H algorithm given appropriate,MCMC method has higher test validity because of making full use of both the likelihood function and the prior information,and has completely accurate judgment in larger samples. Moreover,MCMC method is also very effective for different order spatial adjacency matrix of the spatial econometric model selection.
[1]Burridge,P.On the Cliff-Ord test for spatial autocorrelation[J].Journal of the Royal Statistical Society B,1980,(42):107-108.
[2]Anselin,L.Lagrange multiplier test diagnostics for spatial dependence and spatial heterogeneity[J].Geographical Analysis,1988(20):1-17.
[3]Anselin,Luc,et al.Simple diagnostic tests for spatial dependence[J].Regional Science and Urban Economics,1996(1):77-104.
[4]Hepple,L.W.Bayesian techniques in spatial and network econometrics:Model Comparison and posterior odds[J].Environment and Planning A 27,1995,447-469.
[5]LeSage,J.P.Bayesian estimation of spatial autoregressive models[J].International Regional Science Review,1997(20):113-29.
[6]Jeffreys,H.Theory of Probability[M].Oxford University Press,1961.
[7]Hepple,L.W.Bayesian model choice in spatial econometrics,in Advances in Econometrics:Vol 18:Spatial and Spatiotemporal Econometrics[M].Oxford:Elsevier Ltd,2004.
[8]LeSage,J.P.and Pace R.K.Introduction to spatial econometrics[M].Boca Raton,US:CRC Press Taylor&Francis Group,2009.
[9]Bera A,Yoon M.Simple diagnostic tests for spatial dependence[R].Working paper,Champaign,1992.
[10]Schwarz,Gideon E.Estimating the dimension of a model[J].Annals of Statistics,1978,6(2):461-464.
[11]Burnham,K.P.and Anderson,D.R.Model Selection and Multimodel Inference:A Practical-Theoretic Approach[M].Springer-Verlag,2002.
[12]Carlin B P,Louis T A.Bayes and empirical Bayes methods for data analysis[J].Statistics and Computing,1997 7(2):153-154.
[13]Gelfand,A.E.,Smith,A.F.M.Sampling-based approaches to calculating marginal Densities[J].Journal of the American Statistical Association,1990(85):398-409.
[14]Hastings,W.K.Monte Carlo sampling methods using Markov chains and Their applications[J].Biometrika,1970(57):97-109.
[15]Geweke,John.Bayesian Treatment of the Independent Student t Linear Model[J].Journal of Applied Econometrics,1993(8):19-40.
[16]LeSage,J.P.Bayesian estimation of spatial autoregressive models[J].International Regional Science Review,1997(20):113-129.
1 本文只针对横截面的空间计量模型进行了分析。另外,基于表格大小,文章篇幅及工作量的考虑,本文模拟分析部分并未对10 种模型全部进行模拟,而是对不同的模型选择方法进行了有针对性的模拟分析。
1 在常见软件中,进行空间计量模型的估计时,都没有输出信息准则值,可以根据给出的对数似然函数值利用下面介绍的公式进行计算,本文模拟部分的信息准则值是使用matlab计算得到的。
2 要注意的是Eviews软件中给出的信息准则值是样本平均意义上的信息准则值,详见EViews 8 Users Guide I,Quantile Regression,Chapter 32,(11.89),441页。本文模拟分析部分的信息准则值是根据似然函数值通过matlab2013a手动计算得到的。
1本算法的全部过程由存在异方差与同方差两个迭代分支,为避免叙述重复冗长,仅给出了后者。
1分别为美国俄亥俄州的犯罪数据和1980年总统选举数据中的空间加权矩阵,详见Anselin的Spatial Econometrics:Methods and Models和Pace和Barry的Quick computation of spatial autoregressive estimators,Geographical Analysis。
基本信息:
DOI：10.19343/j.cnki.11-1302/c.2014.08.013
中图分类号:O212
引用信息:
[1]陶长琪,杨海文.空间计量模型选择及其模拟分析[J].统计研究,2014,31(08):88-96.DOI:10.19343/j.cnki.11-1302/c.2014.08.013.
基金信息:
国家自然基金项目(编号:71073073,71273122)的资助
2014-08-15
2014-08-15