用线性模型模拟线性-对数问题产生的自相关情况统计分析

摘要

用线性模型模拟线性-对数问题，残差序列会表现出自相关性。从线性-对数模型的参数取值大小、随机误差项方差大小、样本范围大小等三个方面，对用线性模型模拟线性-对数问题产生的自相关情况进行了统计分析，结果表明，当半对数模型的参数取值较大、随机误差项较小、样本范围较大时，残差序列的自相关性的百分比较大，此时不易误用线性模型。该研究对加深认识线性-对数模型，避免误用线性模型有一定的参考意义。%While the linear-logarithmic problem is simulated by a linear model, the residual error series may produce autocorrelation. In this paper, the statistic analysis of the autocorrelation is achieved in terms of the parameter value, the variance of the random error and the sample range of the linear-logarithmic model. The final statistical analysis indicates: when the semi-logarithmic model parameter is larger, the random error is smaller; when the sample range is broader, then the percentage of autocorrelation of residual error percent is higher. In two cases it is not easy to misuse of linear model. This result has a certain reference significance to deepen the understanding of linear- logarithm model in order to avoid the misuse of linear model.