摘要
本文采用组合模型的形式对时间序列数据的变化特点建模,在模型通过各种检验、具有良好统计预测功能的基础上,从检验异常值的角度来分析预测值与实际值之间差异的程度,找出离群数据,利用数理统计中检验实验观测数据异常值的方法,对离群数据的误差进行统计上的显著检验,从而评估统计数据的质量。文章以我国国内生产总值(GDP)为研究对象,选取我国1978—2003年间的GDP作为样本,运用趋势模拟评估法来评估我国2004年国内生产总值的准确性。对我国经济指标的时间序列数据进行了实证分析。
This paper makes use of combined models to model the change of time series data, after the model passes various statistical test and has the effect of prediction, then analyzes the difference between predicted value and actual value from the angle of outliers to find outliers of data, at last uses the method of mathematical statistics to test the outliers and evaluate the quality of statistical data. After introduction to this method, this paper chooses series of GDP from 1978 to 2003 as a sample and uses this method to evaluate the accuracy of GDP in 2004.
出处
《统计研究》
CSSCI
北大核心
2007年第8期17-21,共5页
Statistical Research
关键词
统计数据质量
趋势模拟评估法
实证分析
Quality of statistical data
Method of trend simulation
Empirical analysis
