Russell Davidson

Several Tests for Model Specification in the Presence of Alternative Hypotheses

Several procedures are proposed for testing the specification of an econometric model when one or more models purport to explain the same phenomenon. These procedures are closely related, although not identical, to non-nested hypothesis tests proposed by Pesaran and Deaton, and have similar asymptotic properties. They are simple conceptually and computationally, and unlike earlier techniques, may be used to test against several alternative models simultaneously. Some empirical results suggest that ability of the tests to reject false hypotheses is likely to be good in practice.

Tests for model specification in the presence of alternative hypotheses: Some further results

Abstract In Davidson and MacKinnon (1981), two of the present authors proposed a novel and very simple procedure for testing the specification of a nonlinear regression model against the evidence provided by a non-nested alternative. In this paper we extend their results in several directions. First, we relax a number of the assumptions of the previous paper; we admit the possibility that the nonlinear regression functions may depend on lagged dependent variables, and we do not require that the error terms be normally distributed. Second, we show how the earlier procedure may straightforwardly be generalized to the case where the two non-nested models involve different transformations of the dependent variable. Finally, we propose a simple procedure for testing non-nested linear regression models which have endogenous variables on the right-hand side, and have therefore been estimated by two-stage least squares.

Distribution-Free Statistical Inference with Lorenz Curves and Income Shares

The paper considers the problem of statistical inference with estimated Lorenz curves and income shares. The full variance-covariance structure of the (asymptotic) normal distribution of a vector of Lorenz curve ordinates is derived and shown to depend only on conditional first and second moments that can be estimated consistently without prior specification of the population density underlying the sample data. Lorenz curves and income shares can thus be used as tools for statistical inference instead of simply as descriptive statistics.

Bootstrap tests: How many bootstraps?

In practice, bootstrap tests must use a finite number of bootstrap samples. This means that the outcome of the test will depend on the sequence of random numbers used to generate the bootstrap samples, and it necessarily results in some loss of power. We examine the extent of this power loss and propose a simple pretest procedure for choosing the number of bootstrap samples so as to minimize experimental randomness. Simulation experiments suggest that this procedure will work very well in practice.

Convenient specification tests for logit and probit models

We propose several Lagrange Multiplier tests of logit and probit models, which may be inexpensively computed by artificial linear regressions. These may be used to test for omitted variables and heteroskedasticity. We argue that one of these tests is likely to have better small-sample properties, supported by several sampling experiments. We also investigate the power of the tests against local alternatives. The analysis is novel because we do not require that the model which generated the data be the alternative against which the null is tested.

Graphical Methods for Investigating the Size and Power of Hypothesis Tests

Simple techniques for the graphical display of simulation evidence concerning the size and power of hypothesis tests are developed and illustrated. Three types of figures--called P value plots, P value discrepancy plots, and size-power curves--are discussed. Some Monte Carlo experiments on the properties of alternative forms of the information matrix test for linear regression models and probit models are used to illustrate these figures. Tests based on the outer-product-of-the-gradient regression generally perform much worse in terms of both size and power than efficient score tests. Copyright 1998 by Blackwell Publishers Ltd and The Victoria University of Manchester

The Size Distortion Of Bootstrap Tests

We provide a theoretical framework in which to study the accuracy of bootstrap P values, which may be based on a parametric or nonparametric bootstrap. In the parametric case, the accuracy of a bootstrap test will depend on the shape of what we call the critical value function. We show that, in many circumstances, the error in rejection probability of a bootstrap test will be one whole order of magnitude smaller than that of the corresponding asymptotic test. We also propose a simulation method for estimating this error that requires the calculation of only two test statistics per replication.

Statistical inference for the measurement of the incidence of taxes and transfers

We establish the asymptotic sampling distribution of general functions of quantile-based estimators computed from samples that are not necessarily independent. The results provide the statistical framework within which to assess the progressivity of taxes and benefits, their horizontal inequity, and the change in the inequality of income which they cause. By the same token, these findings characterise the sampling distribution of a number of popular indices of progressivity, horizontal inequity, and redistribution. They can also be used to assess welfare and inequality changes using panel data, and to assess poverty when it depends on estimated population quantiles. We illustrate these results using micro data on the incidence of taxes and benefits in Canada.

Some Non-Nested Hypothesis Tests and the Relations Among Them

This paper discusses several statistical techniques which can be used to test the validity of a possibly nonlinear and multivariate regression model, using the information provided by estimating one or more alternative models on the same set of data. The techniques we propose can be regarded as alternative implementations of Coxs idea for non-nested hypothesis testing; under the null hypothesis, all of the test statistics are asymptotically the same random variable. For the univariate linear regression case, our test and Pesarans has asymptotic relative efficiency of unity for local alternatives. Finally, we present sampling experiments for univariate linear models which show that the small-sample performance of our J test and Pesarans test can be quite different.

Testing for Restricted Stochastic Dominance

Asymptotic and bootstrap tests are studied for testing whether there is a relation of stochastic dominance between two distributions. These tests have a null hypothesis of nondominance, with the advantage that, if this null is rejected, then all that is left is dominance. This also leads us to define and focus on restricted stochastic dominance, the only empirically useful form of dominance relation that we can seek to infer in many settings. One testing procedure that we consider is based on an empirical likelihood ratio. The computations necessary for obtaining a test statistic also provide estimates of the distributions under study that satisfy the null hypothesis, on the frontier between dominance and nondominance. These estimates can be used to perform bootstrap tests that can turn out to provide much improved reliability of inference compared with the asymptotic tests so far proposed in the literature.