期刊名称:Case Studies in Business, Industry and Government Statistics
印刷版ISSN:2152-372X
出版年度:2008
卷号:2
期号:1
页码:1-8
出版社:Bentley University
摘要:Given a time series that reacts to an intervention, this article illustrates two points: (1) how simulations, either based on a normality assumption or through bootstrapping, can help us measure the impact of the intervention not necessarily on the mean but on functionals of the time series; and (2) how backcasting the series can help finding the time interval necessary for the series to recover a regular dynamic after the intervention. The methodology has been developed using an ARIMA model, but could equivalently be based on alternative models such as basic structural models, or on exponential smoothing.?