Prediction Tests for Structural Stability of Multiple Time Series
研究基于预测与实际值比较来检验多元时间序列结构稳定性的方法，发现单变量检验与多变量检验各有优势，建议实践中同时使用两种检验，并用西德经济数据验证了方法在石油危机期间的有效性。
Comparing forecasts with actually observed values is one possible way to check the structural stability of time series. In this study, prediction tests based on this principle are investigated in the context of multiple time series analysis. It is found that the power of tests based on univariate series may exceed that of tests based on multivariate series and vice versa. Therefore, it is proposed to use both kinds of tests in practice. The tests are used to investigate the stability of a system of West German economic data during and after the 1973–1974 and 1979–1980 oil price increases.