Investigation of Two Preliminary Analysis-Altering Elements: Ordinate Scaling and DPPXYR.
Squeezing the y-axis or hiding data-point boxes does not fool BCBAs into seeing stronger effects.
01Research in Context
What this study did
Peltier et al. (2024) asked 60 BCBAs to look at ABAB graphs.
They changed two things: they squeezed the y-axis (ordinate scaling) and removed the little boxes around each data point (DPPXYR).
Analysts rated how sure they were that the treatment worked and how big the effect looked.
What they found
Shrinking the y-axis did not make analysts more confident.
Taking away the data-point boxes also did nothing for confidence.
Both tweaks gave mixed results on how large the effect seemed.
How this fits with other research
Dowdy et al. (2024) ran a similar 2024 test and found that changing the x-to-y axis ratio did sway visual judgments.
The two studies look opposite, but Corey tested y-axis only while Dowdy tested the ratio—so the tweaks hit different visual cues.
Wolfe et al. (2023) showed that trend and variability inside the data matter more than layout tricks.
McGonigle et al. (1982) already told us experts watch effect patterns, not tiny graph details.
Together, the papers say: data shape beats window dressing.
Why it matters
You can stop fiddling with y-axis limits or fancy point markers. Focus on clean data and clear trends instead. Your clinical judgment will stay steady no matter how the graph is dressed up.
Want CEUs on This Topic?
The ABA Clubhouse has 60+ free CEUs — live every Wednesday. Ethics, supervision & clinical topics.
Join Free →Leave your y-axis auto-scaled and keep the data-point boxes—then spend that saved time checking for trend and variability instead.
02At a glance
03Original abstract
The purpose of this pre-registered study (Peltier & McKenna) was to conceptually replicate if the truncation of the ordinate and DPPXYR increased analysts' estimation of a functional relation and magnitude of treatment effect. Visual analysts (n = 27) evaluated eight data sets reporting null (n = 2), small (n = 2), moderate (n = 2), and large (n = 2) effects. Each data set was graphed six times with manipulations of the ordinate and DPPXYR, resulting in 48 ABAB graphs. We estimated two separate three-level mixed effect models with variations nested in datasets and nested in participants to evaluate the impact of graph characteristics for (1) confidence in determining a functional relation and (2) the estimated magnitude of the treatment effect. We included ordinate scaling and DPPXYR at level 1 and graph effect size at level 2, including all interactions. Overall, graph manipulation consistently did not impact confidence in a functional relation. Results suggest mixed findings for graph manipulation on the estimated magnitude of the treatment effect. Findings will be couched in current literature and recommendations for graph construction and future research will be discussed.
Behavior modification, 2024 · doi:10.1177/01454455231221289