Building Consumer Understanding by Utilizing a Bayesian Hierarchical Structure within the Behavioral Perspective Model.

Rogers et al. (2017) wrote a how-to paper. They showed how to plug Bayesian hierarchical modeling into the Behavioral Perspective Model (BPM).

The BPM tracks how price, reward, and consumer history shape buying choices. The authors added a statistical layer that updates each shopper’s personal weights as data arrive.

The paper is a recipe, not an experiment. No new data were collected, so no outcome is reported.

The authors simply show that Bayesian estimates of BPM parameters are more stable than single-level fits when choices are noisy or sample sizes are small.

Franck et al. (2019) extends the same idea to delay-discounting data. They swap consumer choices for money-now versus money-later and still gain cleaner parameter estimates with Bayesian hierarchies.

Alaimo et al. (2015) is the predecessor. They first used Bayesian model selection to pick the best discount curve per person. Rogers et al. (2017) borrow that logic and move it into the BPM world.

Killeen (2019) joins the chorus against p-values. All four papers agree: Bayesian credible intervals beat null-hypothesis tests when you want to predict, control, and replicate behavior.

If you analyze choice data—shopping, food selection, or token economies—this paper gives you a ready script. Fit the BPM in a Bayesian hierarchy and you get tighter estimates of reinforcer value, even with small caseloads. Share the code with your data team and start reporting credible intervals instead of p-values.