优化STR模型的实证研究

Empirical Research on Optimized STR Model

采用平滑转换回归(smooth transition regression,STR)模型对GDP与工资总额之间的非线性关系进行了研究。通过各种检验后选择LSTR模型进行回归分析,得到了GDP与工资总额之间的非线性函数关系。在采用高斯-牛顿算法获得参数估计的基础上,应用搜索法优化LSTR模型参数的估计。结果表明,用搜索法对参数进行估计,优化的STR更具有准确性,可提高模型的拟合程度。

The smooth transition regression （STR） model was used to analyze the inherent relationship between GDP and to- tal wages. The LSTR model was chosen to make regression analysis after a variety of hypothesis testings. Then the nonlinear rela- tionship between the GDP and total wages was obtained. On the basis of achieving parameter estimation by Gaussian - Newton al- gorithm, Grid search method was used to optimize the estimation of LSTR model parameter. The results show that the Grid search method is better than Gaussian - Newton algorithm in parameter estimation.