We introduce a new method to robustifying inference that can be applied in any situation where a parametric likelihood is available. The key feature is that data from the postulated parametric models are assumed to be measured with error where the measurement error distribution is chosen to...