确定时间序列均值和方差函数的移动多点平均方法

为了提高有限样本或小样本情况下时间序列均值和方差函数的确定精度,以保证时序建模、分析和预测精度,提出一种确定序列趋势项的移动多点平均方法.该方法能够得到时间序列均值中非周期部分,结合样本周期图法得到的周期项,可综合得到其均值函数,并可进一步得到时序的标准差函数.MonteCarlo模拟对比结果表明:该方法不仅能够保证时间序列段内分析精度,而且能够有效提高时间序列的预报精度.

Moving multi-point average method for determining functions of time series mean and variance

A moving multiple-point average (MMPA) method was proposed to enhance the determining precision for time series mean and variance functions, and consequently to improve the accuracy for time series modeling, analyzing and forecasting. Non-periodic part of trend item which can be obtained by the MMPA approach, and the periodic part of trend item which can be calculated by sample periodogram method, comprise the whole trend function. We can further determine the series standard deviation function. Monte Carlo simulation study shows that not only the analysis precision but also the prediction accuracy can be ensured by the established methodology.