Non-nested models and the likelihood ratio statistic: A comparison of simulation and bootstrap based tests

摘要: We consider an alternative use of simulation in the context using Likelihood-Ratio statistic to test non-nested models. To date has been used estimate Kullback-Leibler measure closeness between two densities, which turn 'mean adjusts' statistic. Given that this adjustment is still based upon asymptotic arguments, procedure utilise bootstrap procedures construct empirical density. our knowledge study represents first comparison properties and simulation-based tests applied tests. More specifically, design experiments allows us comment on relative performance these testing frameworks across models with varying degrees nonlinearity. In respect although primary focus paper evaluation simulationand bootstrap-based nonnested a class nonlinear threshold models, inclusion similar analysis more standard linear/log-linear provides point comparison.