摘要
误差校正模型具有较好的预测能力,在时间序列分析中占据重要地位。将误差校正模型从均值框架推广到分位数框架,提出了分位数误差校正模型的概念,并给出一整套建模技术:模型表示、参数估计、模型定阶、诊断检验、密度预测等。通过数值模拟,将其与经典的均值误差校正模型、分位数自回归模型进行比较,发现分位数误差校正模型极大地提高了预测的准度与精度。此外,选取中国货币供应与物价水平之间关系作为研究对象,实证检验了分位数误差校正模型的条件密度预测能力。
Error correction model plays an important role in time series analysis, as it performs good in forecasting. We propose the concept of quantile error correction model (QECM) by extending the error correction model from mean framework (MECM) to quan- tile framework. The methods for QECM, include model structure, parameter estimation, order identification, diagnostic test and density forecasting. QECM is proved to be superior to MECM and quantile autoregressive model in terms of the accuracy and precision of fore- casting through numerical simulations. The empirical analysis on the relationship between money supply and price level also confirms that QECM has good performance in density fore- casting.
出处
《数量经济技术经济研究》
CSSCI
北大核心
2016年第10期110-127,共18页
Journal of Quantitative & Technological Economics
基金
国家自然科学基金(71071087)
教育部人文社会科学研究规划基金项目(14YJA790015)
安徽省哲学社会科学规划基金项目(AHSKY2014D103)
合肥工业大学产业转移与创新发展研究中心招标项目(SK2014A073)的资助
关键词
分位数回归
误差校正
条件概率
预测精度
Quantile Regression
Error Correction
Conditional Probability~ Forecas- ting Accuracy