噪音比之檢定

Abstract

This paper recommends the Wald test to test the “adjusted noise-to-signal ratio”. Kaminsky, Lizondo and Reinhart (1998) use “adjusted noise-to-signal ratio” to examine the effectiveness of individual indicators. However, they did not provide a formal statistical test to test whether “adjusted noise-to-signal ratio” is equal to 1. By “adjusted noise-to-signal ratio” concept, we can measure the accuracy of sequence of event forecasts. In this paper the Wald test for testing “adjusted noise-to-signal ratio” can derived under i.i.d. and Markov chain DGP’s. Besides, the sizes of the Wald test suffer very little sign of distortion. Therefore we recommend the Wald test to test the“adjusted noise-to-signal ratio”. We also provide an empirical example in the end of this paper. The empirical data are obtained from Kaminsky and Reinhart (1996) paper, and we use the Wald test to test the “adjusted noise-to-signal ratio” of indicators. We find that 15 out of the 16 indicators are effective. Only the indicator of real interest differential is unuseful.