Modelling Non-Linear Relationships Between Long-Memory Variables

点击翻译按钮获取中文摘要

This project develops new procedures for estimating non-linear models of "long-memory" processes, that is relationships between values over long periods. These new procedures can be applied to many questions in macro, financial and labor economics. They are particularly relevant for modelling relationships between macroeconomic and weather variables, as when studying global warming for instance, as these variables are certainly slowly moving and the relationships are likely to be non-linear. Work concerning long memory processes has largely been linear. There are many non-linear modelling techniques available to approximate actual relationships but these procedures usually assume, either explicitly or implicitly, that the variables are stationary, or short-memory. In order to provide a class of techniques appropriate for modelling non-linear relationships between long variables, this project will develop new definitions and tests of long-memory in the non-linear framework and specify new techniques for testing for linearity and for overcoming the "data-missing" properties of some of the techniques.