How do we read scatterplots?

Abstract

People look for patterns, structures, traits, trends, anomalies, and correlations in data. Data visualization helps with this by presenting the data in various formats with various interactions. It can give an qualitative perspective of huge and complex data sets. Additionally, it can provide a data summary, help identify areas of interest, and suggests acceptable parameters for more specialized quantitative research. The scatterplot is arguably the most popular data display method which makes it easier to identify clusters, trends, and correlations. However, they can quickly become too overloaded from the user's perspective when there is a lot of data available. Overplotting is a problem that occurs when multiple observations (points) have the same or strikingly similar values, making it difficult for the user to understand the relationships between the points and variables and producing inaccurate or misleading information in the graph. In this study, we analyze how the size of data points affect the perception of regression in overloaded scatterplots. Furthermore, we analyze if the education and/or experience in data visualization affects the perception as well. In addition to adhering to the fundamentals of quantitative research by introducing various types, assumptions, techniques, and common mistakes that many researchers make when conducting research studies, this study is dependent on the fundamental and practical issues that should be taken into account when pursuing evaluation studies in information visualization. Our results show that increasing the dot size does have a positive effect in recognizing the regression in a overplotted scatterplot. Even if the individual dots are not visible anymore the amount of people who see the regression correctly increase. Furthermore, the results show that experience in data visualization does not affect the recognition of regression. Education, however, may affect the recognition of it.