关键词: Bootstrap抽样, Moran’s I检验, 空间面板数据, Monte Carlo模拟

Abstract: Moran’s I test has been widely used for spatial autocorrelation. When the error term is non-normally distributed, Moran’s I test turns out to be large size distortion, leading to a poor performance of the test. This study adopts the modified Bootstrap methods to improve the performance of Moran’s I test. Monte Carlo simulation experiment results suggest that, when the error term is normally distributed, both asymptotic Moran’s I test and Bootstrap Moran’s I test have superior size and power performance; when error term is serial correlated, asymptotic Moran’s I test results in a serious size distortion, while bootstrap Moran’s I test has a better performance, which has effectively reduced the size distortion and increased the power of asymptotic Moran’s I test.

Key words: Bootstrap Method, Moran’s I Test, Spatial Panel Data, Monte-Carlo Simulation