Post date: Dec 17, 2015 10:27:29 PM
In a meeting with Jason O'Rawe today, we were talking about how similar the c-box approach and (robust) Bayesian methods are, and we at first jokingly said we should re-brand c-boxes as "Quick Bayes". But after laughing a bit, we started to realize that this might actually be reasonable. It is arguably at least as reasonable as using the word "Bayes(ian)" in the phrases "Bayesian model averaging" or "empirical Bayes", and maybe "approximate Bayesian computation".
It's quick because (1) it's computationally cheaper in many situations, (2) you don't have to select a prior, and (3) the sensitivity analysis is already built into the calculation.
It's Bayes because (1) often agrees with Bayes in the one-dimensional case with Jeffreys priors, (2) generally encloses results from multiple uninformative priors that have been suggested, and (3) can be generalized to be compatible with Bayesian methods by specifying a prior or class of priors.
Given the obvious benefits in the marketplace of ideas, we are seriously considering re-branding ourselves.