使用時間數列分析商業及經濟資料時，對於如何處理各式各樣的干擾，諸如政策更異，能源危機，罷工，以及促銷等，通常須格外注意。在本文中，吾人採用返覆程序偵測時間數列中的可能干擾。本文所提之方法適用於單元自迴歸累積移動平均模式 (autoregressive integrated moving-average, ARIMA (p,d,q), model) 及動態迴歸模式 (dynamic regression model)。至於所考慮之特異或干擾效應包括創新性特異，增減性特異，水平移位，以及暫時性變更。吾人採用一般動態迴歸模式來研究是各種干擾效應，並且利用返覆程序來處理模式參數及干擾效應之聯合估計。此外，吾人亦研究特異及干擾效應對參數估計及預測之影響。文中並分析二則實例，包括一套臺灣能源資料，以資佐證。

Time series analysis of business and economic data often requires special attention in order to handle various interventions such as economic policy change, energy crisis, strike, and sales promotion. In this paper, we consider an iterative procedure to detect possible interventions of a time series. The process studied may follow a univariate ARIMA or a dynamic regression model, and the intervention effects enter tained include additive and innovation outliers, level shift, and temporary change. A general dynamic model is employed to handle the effects of detected interventions. An iterative procedure for the joint estimation of model parameters and intervention effects is considered. We also discuss briefly the effects of interventions on parameter estimation and forecasting of a time series. Two illustrative examples including an economic data set of Taiwan are given.