具有AR(p)误差的非线性模型异方差和相关性检验

Testing for heteroscedasticity and autocorrelation in nonlinear models with AR(p) errors

讨论随机误差是AR(p)序列的非线性回归模型的异方差和自相关性检验问题.首先导出联合检验的score统计量,然后利用参数的正交变换,得到了调整的score检验统计量.当模型存在自相关性时,给出了检验异方差性的score统计量和调整的score统计量.最后利用得到的检验方法分析了氯化物数据,分析结果表明,该数据具有显著的异方差和AR(2)相关性.

The tests for heteroscedasticity of random errors and autocorrelation in nonlinear regression models with AR （p） errors are discussed. Firstly, the score test statistic for composite test of heteroscedasticity and autocorrelation is developed. Then, based on the parameter orthogonality transformation, the modified score test is derived. In the autocorrelated models, the score test statistic and its adjustment are given. The chloride data is used to illustrate our results, which shows that these data have significant heteroscedasticity and AR（2） autocorrelation.