摘要
讨论误差服从一阶自回归的线性模型关于回归系数的线性假设检验问题.在方差参数已知的情况下,研究了检验统计量的性质;在方差参数未知时,采用前三阶中心矩相等的方法,提出两种近似检验方法.模拟结果显示,当误差正相关程度较高时,两种近似检验在具有较小的第一类错误的同时,具有较大的功效.
The problem of testing linear hypothesis about regression coefficients of linear model with AR(1) errors is discussed. The properties of tests statistic are studied when the variance parameter is known. For unknown variance parameter case, two approximate tests statistics are proposed by“matching the first three moments”. The simulation results show that for high positive errors correlation cases,these two approximate tests statistics have lower type-1 error and larger power.
出处
《上海理工大学学报》
EI
CAS
北大核心
2005年第4期295-299,共5页
Journal of University of Shanghai For Science and Technology