摘要：This paper attempts to introduce and apply some methods of model selection in order to avoid a specification error. In general, the specification error can be indicated by three possibilities which are possible problems with: (1) disturbances, (2) explanatory variable and (3) parameters. In this paper, our discussion is concentrated on the functional form of our estimable model. The estimated model in this paper is the demand for real currency in Indonesia for the period of 1984(1) - 1997 (IV). The empirical result show that MacKinnon, White and Davidson (MWD Test), Bera and Me A leer (B-M Test) and Zarembka tests are fail to identify the best function at form of our empirical model. However, using error correction approach, the findings show the best functional form of the demand for real currency in Indonesia is the log linear regression mode.
关键词：methods of model selection; functional form and error correction approach
