摘要
本文研究固定效应设定下,空间误差分量(Spatial Error Components,SEC)模型的空间相关性边际检验、条件检验和转换检验.Monte Carlo模拟实验证明,转换检验有更小的水平扭曲和优越的检验功效,且不受固定效应大小影响,是经济计量实证中理想的检验统计量;转换检验有限样本性质受空间权重矩阵的选取和N、T影响较小;选取Queen型矩阵或更大的样本量将增大边际检验的功效,但会导致更大的水平扭曲.
We studied spatial correlation tests of fixed effect spatial error components (SEC) model, including marginal tests, conditional tests and transformed tests. Monte Carlo simulation experiments prove that transformed tests have smaller level tortuosity and predominant power; what's more, they will not be affected by fixed effect, and these advantages make them perfect tests in practical econometric analysis. The small sample performance of transformed tests will be affected slightly by spatial weight matrixes and N/T. Queen-Type matrixes and larger sample will increase power of marginal tests, but they lead to larger level tortuosity.
出处
《经济数学》
北大核心
2011年第4期79-85,共7页
Journal of Quantitative Economics
基金
国家自然科学基金项目(71071060)