Discounting: A practical guide to multilevel analysis of choice data

Young (2018) wrote a how-to guide for researchers who run delay-discounting tasks. The paper shows each step for using multilevel logistic regression instead of older curve-fitting methods.

The guide keeps reward size and wait time in separate model terms. This lets you see which factor really drives the choice pattern.

The article does not give new data. It shows that the old one-curve approach can hide effects. The multilevel model keeps the data nested by person and by trial, so you get cleaner answers.

Moeyaert et al. (2020) use the same multilevel trick, but for meta-analysis of single-case graphs. Both papers walk you through code, so you can copy and paste.

Neely et al. (2024) extend the idea into the big-data age. They tell you to clean your session logs first, then run the same kind of model Young shows.

Hatfield et al. (2019) warn about a different pitfall: testing only within one group. Young avoids that trap by forcing a between-factor structure inside the logistic model.

If you study choice or self-control, swap your old discounting script for a multilevel logistic setup. You will keep magnitude and delay effects apart, spot outliers faster, and meet newer journal stats rules without extra math classes.