Approximate Confidence Distribution Computing (ACDC) offers a new perspective on the rapidly evolving field of improbable reasoning from within a frequentist framework. The attractiveness of this computational method for statistical inference rests on the concept of the confidence distribution, a special type of estimator defined in terms of the repeated sampling principle. The ACDC method provides frequentist verification of computational reasoning in problems with unknown or intractable probabilities. The main theoretical contribution of this work is the identification of congruent conditions required for the frequentist validity of inferences from this method. In addition to providing examples of how to connect Bayesian and frequentist inference paradigms using our most recent understanding of confidence distribution theory, we extend the current scope of so-called approximate Bayesian inference by targeting We present the case of including non-Bayesian inference in It is the confidence distribution, not the posterior distribution. The main practical contribution of this work is the development of data-driven approaches to drive his ACDC in both Bayesian or frequentist contexts. The ACDC algorithm is made data-driven by the choice of data-dependent proposal functions, and its structure is very general and adaptable to many settings. We examine two numerical examples that validate the theoretical arguments in the development of ACDC and suggest examples in which ACDC exerts computational superiority over approximate Bayesian calculus methods.
