A Direct Test for Changing Trend
提出一种无需预先知道变化点位置的趋势变化检验方法，利用布朗桥的击中概率推导临界值，并应用于二战前实际GNP、人均GNP和实际工资数据，拒绝趋势平稳假设。
Abstract We consider tests for changing trend that do not require prior knowledge about the location of the changepoint. The limiting distribution is derived from the functional central limit theorem and the critical value from the hitting probability of a Brownian bridge. Applying a test sensitive to the alternative of trend stationarity with structural breaks, we find that for real gross national product, real per capita gross national product, and real wages before World War II the hypothesis of trend stationarity is rejected. KEY WORDS: Brownian bridgeFunctional central limit theoremHitting probabilityStructural changeTrend stationarity