春节模型的效应期识别方法研究

春节模型的参数设定目前更多依赖于研究者的主观判断.为解决这一问题,本文首先介绍了春节因素调整的一般过程以及模型设定和检验的相关细节.在此基础上,提出一种基于"循环遍历"方式和序贯检验方法来自动选择春节模型的最优参数组合的新思路.并以社会消费品零售总额月度序列的春节模型的最优效应期长度识别为例,检验该方法的可行性和有效性.实证结果显示,基于季节峰值、离群值点、Q统计量、AICC值和BIC值等统计量的序贯检验能够有效识别出春节模型的最优效应期长度,进而改善春节模型的季节调整性能.

Research on the influence period length identification of the Spring Festival model

Nowadays, the parameter setting of the Spring Festival model mainly depends on subjective judgement for researchers. To solve the problem, firstly, the paper introduces factor-adjusted general process on Spring Festival and relevant details for model specification and model inspection. On this basis, the paper puts forward new ideas on selecting automatically Spring Festival model with optimal parameters based on "loop through" and sequential test method. Taking Spring Festival model's optimal effective phase length recognition of total retail sales of social consumer goods as an example, we examine its feasibility and validity. The empirical results show that the sequential test method based on seasonal peaks, outliers, Q statistics, AICC values, and BIC values can effectively identify the optimal infulence period length of the Spring Festival model, which can efficiently improve seasonal adjustment performance on Spring Festival model.