Search

Graphical statistics plays a very important role in research and industry. As a statistician, it is very useful to know how analysts make decisions based on plots, so that the plots can be improved for better performance. With the help of eye-tracking equipment, researchers can assess this. This paper describes the use of an eye-tracker for assessing how people look at lineup plots that are used in visual inference. The experiment included a comparison of several plot types, histogram, density, scatterplot and dotplot for two group data, and the use of side-by-side boxplots and scatterplots for assessing the importance of predictors in a regression model. Results are compared with those from a larger visual inference study using Amazon's Mechanical Turk. The eye-tracker can be manageably used only on a small sample of people, to give detailed information plot about reading behavior, so it is ideally used prior to a large human subjects study to help refine the experimental procedures, or as a post-processing step to better understand any surprises in the results. [ABSTRACT FROM AUTHOR]