Introduction to rethinking learners' reasoning with nontraditional data

Abstract

Traditional statistics education has focused on data from random samples and has capitalized on knowledge about a sample to understand an unknown population. However, many ubiquitous forms of data in the modern world do not clearly fit the sample-population assumptions that underpin statistical reasoning. For example, data collected in real time (eg, GPS, live traffic, tweets), image based (eg, photographs, drawings, facial recognition), semistructured (eg, data scraped from social media posts), repurposed (eg, school testing data to estimate housing prices), and big data (open access internet data, civic databases) are all examples of nontraditional data. Traditional data and data sources are typically simpler in nature, static, and more structured, whereas nontraditional forms of data are large, messy, complex, semistructured, or unstructured, constantly changing or evolving, and come in many different formats. While nontraditional forms of data and reasoning about uncertainty have been with us for some time [3,7,10], the digital age has led to a pervasive culture of data in all aspects of life, including those of our students. Widespread availability and access to myriad nonconventional, repurposed, massive, or messy data sets necessitate broadening educational knowledge to better understand how learners make sense of, model, analyze, and make predictions from these data. New research directions have emerged, focusing on methods for making predictions from open, semirelated, and ubiquitous data, often relying heavily on computational methods and predictive modeling. Concerns have been expressed about the relative lack of attention to how and why data were collected, whether inferences being made are trustworthy and how statistics education might respond (eg, [18]). We are united in our goal to develop learners' deep understanding and reasoning with data and models. Therefore, awareness of the implications of nontraditional data—including complexities resulting from the contexts in which data are generated—has resulted in multiple discussions about how the field of statistics education may proceed (eg, [1,4,6,11,13]), but many questions remain open. This special issue addresses some of the open questions in how the field of statistics education may begin to support the teaching and learning of methods for working with nontraditional data. The articles in this issue focus on new approaches to the teaching and learning of data practices related to messy, complex, or nontraditional data from the youngest learners [8,19] to secondary learners [14,17], undergraduate students [15,16], graduate students, teachers, and researchers [2,5,9,17]. There are two overarching themes in the articles in this special issue: new ways to consider data visualizations in the classroom [2,5,14,17,19] and new approaches or elements that need to be considered in the teaching and learning of data science practices [8,9,15,16].