Testing for deterministic trend and seasonal components in time series models

Abstract

SUMMARY A univariate time series model can be set up as the sum of trend, seasonal and irregular components. The trend and seasonal components will normally be stochastic, but deterministic components arise as a special case. This paper develops a test that the trend and seasonal components are deterministic using the approach of Lehmann. The procedure is then extended to test for deterministic components in a model formulated in first differences. Both tests are exact and critical values are tabulated.