The natural mathematics of behavior analysis

Li et al. (2018) wrote a how-to guide for testing old operant models. They used VI schedule data and checked many measures at once. The tool is called MCMC-ABC, a kind of smart computer search.

They wanted to see where classic models break. Instead of eye-balling graphs, they let the computer hunt for mismatches.

The model fit some parts of the data but missed others. The paper shows the exact spots where the fit goes bad. The Bayesian search made the flaws easy to see.

Li et al. (2018) pairs with Li et al. (2018) on ETBD. Both papers use multivariate checks on VI data. One tests Catania’s Operant Reserve, the other tests McDowell’s ETBD. Same method, different theory.

Morris et al. (2021) took the ETBD idea and aimed it at clinical data. They modeled self-injury subtypes instead of lever presses. The 2018 math stays the same; only the target behavior changes.

Hake et al. (1983) asked analysts to borrow new lab tools. Li et al. (2018) answers that call by handing over MCMC-ABC. The 1983 wish list finally gets a 2018 tool.

You no longer have to trust a model just because the curve looks close. Run MCMC-ABC on your own schedule data to see where the model slips. If the fit fails, you know the theory needs work before you use it in treatment. That keeps your decisions honest and your graphs truthful.