Research On The Efficiency Of Bootstrap LM Tests In Spatial Panel Data Fixed Models

Adopting asymptotic LM test which based on error term obeys to normal independent identically distributed and sample is large to judge spatial correlation in Spatial econometric analysis of the specific model.However,in the empirical research,the sample is limited or error term does not meet asymptotic LM test assumptions,leading to asymptotic LM test of spatial correlation failure.Adopting FDB method and constructing the Bootstrap LM statistics,through a large number of Monte Carlo simulation from two aspects: size distortion and power to prove the Bootstrap LM test effectively and solve the poor performance of asymptotic LM test,also prove the researcher in empirical research of spatial econometric analysis with effective analysis tool.The paper is divided into three parts.The first part is the introduction and literature review,based on the research background and significance and combing the literature,confirms the research content and research purpose and the technical route of research;The second part describes the idea of Bootstrap LM test research and Monte Carlo simulation in spatial panel data fixed effects models;The third part,analysis the simulation result of the Bootstrap LM and draws the conclusion.The Monte Carlo experiments show that both the asymptotic LM test and Bootstrap LM test work well in finite samples when the error is normally distributed.When the error term has heteroscedasticity or time series related,asymptotic LM test exists serious size distortion and leads to failure in the same condition,the Bootstrap LM test without losing the power of the asymptotic LM test has effectively rectified the size distortion.Thus,Bootstrap LM test is a better test method.In short,whether error term satisfy normal independent identical distribution,the performance of the Bootstrap LM test is better than the asymptotic LM test.So,the Bootstrap LM test is effective inspection method.