Abstract:
Recently, there has been a growing interest in integrating data visualization research with basic visual perception findings. Following this line of reasoning, in this thesis, I investigated how our understanding of ensemble perception, our visual system’s ability to accurately and rapidly extract summary information of briefly presented set of objects that are spatially or featural similar, can contribute to our understanding of scatterplot processing. Across two experiments, I sought to answer two separate yet related questions. One, I investigated whether the presence of an outlier could influence how viewers extract best-fits in scatterplots. Two, I investigated whether familiar content and the presence of trend-consistent outliers influenced best-fits extracted in scatterplots. In both experiments, I briefly presented participants with scatterplots that varied outlier presence. Then, participants drew their best-fit estimates by using the mouse. Comparing their responses with possible best-fit alternatives, I found that outliers are equally weighed with the remaining points in trend-line estimates when there was not any context. However, when the relationship depicted in the scatterplot was familiar, and the outlier point represented a trendconsistent position, viewers were more likely to generate best-fits that overweighed those outliers in their responses, demonstrating that prior beliefs could influence our trend-line estimates from briefly presented scatterplots.
