Relations Among Some Non-Nested Hypothesis Tests

This paper discusses statistical techniques to test the validity of a possibly nonlinear and multivariate regression model, using the information by estimating one or more alternatives on the same data set. The procedures suggested by Pesaran and Deaton (1978) are not the only ones which may legitimately implement Cox's (1961) basic idea of non-nested hypothesis testing. Several other procedures are easier to perform, especially in the multivariate case. Empirical results for demand system models suggest these tests are abundantly capable of rejecting false models, but that alternative tests can yield quite different results.