How Transformations Affect Visual Judgments in Time Series Graphs

There is a Poster PDF for this presentation, but you must be a current member or registered to attend AVS 2026 to view it.
Please go to your Account Home page to register.

Bernice Rogowitz¹ (bernicerogowitz@yahoo.com), Paul Borrel; ¹Visual Perspectives Research, ²Independent Researcher

It is well known that visual representations can profoundly influence how data are perceived. To tease out the visual factors contributing to these judgments we created 48 visualizations of the same time-series COVID-19 mortality dataset. We introduce the distinction between data transformations, which plot transformed values of the data, and rendering transformations, which plot the original data values on transformed axes. We covaried three data transformations: Normalization, Accumulation, and Time-range Subsetting and three rendering transformations: Logarithmic, Context, and Y-axis Scaling. Graphs showed data for France alone, or alongside data from the US and Peru, which differed in mortality and population. Observers rated how safe they would feel traveling to France. Since all graphs displayed the exact same data, the null hypothesis would predict that they would be judged equally. This hypothesis was not supported: for all subjects, rated safety was influenced by these transformations. For this data set: (1) Rendering transformations had a greater effect on judgments than data transformations. What the data looked like on the page mattered more than the numbers themselves. (2) Linear representations were judged to be safer than logarithmic (p=.001). (3) Providing the context of other countries increased perceived safety (p=.0001). To examine visual factors, we also characterized the visual shape and appearance of each graph. Across all graph shapes, the final y value of the curve was the strongest predictor of perceived safety (r2= .58; p<.001). The higher the final mortality value on the graph, the lower the perceived safety, regardless of actual data values. These effects held across all participants; even observers with strong mathematical and visualization expertise were influenced by rendering transformations. These results provide insight into how seemingly neutral visualization decisions affect perceptual judgments and data understanding, which can have important implications in diverse application domains.

Thank You to Our AVS 2026 Sponsors