ADF檢驗中滯后長度的選擇——基于ARIMA

ADF檢驗中滯后長度的選擇——基于ARIMA(0,1,q)過程的模擬證據(jù)

【摘要】在進行ADF檢驗時如何確定一個最優(yōu)的滯后長度一直是研究者們關注的問題。最近的研究表明，不同的滯后長度選擇方法對ADF檢驗的統(tǒng)計推斷影響很大。本文在已有研究的基礎上，模擬了更為一般的ARIMA(0,1,q)過程，分析了在不同的數(shù)據(jù)生成過程、檢驗式以及樣本容量下，各種滯后長度選擇方法對ADF檢驗功效和實際檢驗水平的影響，最后認為修正的信息準則通常具有較合理的實際檢驗水平，而從一般到特殊法具有更為穩(wěn)健的ADF檢驗性質。

關鍵詞 ADF檢驗 滯后長度 信息準則 修正的信息準則 從一般到特殊法

Abstract: The optimal lag length in estimating Augmented Dickey-Fuller statistics have been concentrated on for years. Previous research indicated that different leg length selection models affect a lot on the statistical inference of ADF test. Based on all the researches available, this paper simulates a more general ARIMA(0,1,q) process and analyzes the influence of lag length selection criterions to the size and power of the ADF test with different data generating processes, ADF regressions, and sample sizes. Finally, it is proved that the Modified Information Criteria always shows a more proper size and the General to Special Criteria has more robust properties in ADF test.

Keywords: ADF test Lag Length Information Criteria Modified Information Criteria General to Specific