An analysis-of-variance model for the intrasubject replication design.

Christophersen et al. (1972) wrote a how-to paper. They showed that one- or two-way ANOVA can test reversal-design data.

The model works for one kid or many kids. It gives F scores and p values instead of just eyeballing lines.

The math lines up with visual peaks and drops. ANOVA can flag a real effect even when the graph looks messy.

Diller et al. (2016) and Wolfe et al. (2023) prove the problem R et al. saw: BCBAs often disagree when they eyeball graphs. Trend and scatter make raters split.

Lloveras et al. (2025) offer a newer fix. Their compound multiple-baseline design skips the need for the ANOVA patch by adding a second stagger.

Iversen (2021) pushes back. He says group stats hide individual quirks, so stick with visual logic. The papers do not clash—they warn against blind number-crunching.

You now have two tools for messy reversal graphs: ANOVA or a second stagger. If the baseline drifts, try Lloveras-style double staggering first. If you must stay with classic ABAB, run the ANOVA to check your visual call, but still plot each participant alone so you do not miss individual quirks.