Anna Fergusson

Cleaning data is an important facet of statistical practice. The research literature on examining data practices of learners when dealing with messy data that needs cleaning, however, is scarce. As part of a larger study, six Grade 12 high school statistics teachers engaged with a height estimation task, for which the data were drawn from a publicly available website containing 39,195 rows of text entries in a variety of measurement systems. The teachers’ observed data practices were characterised as inspecting, ideating, sorting, sampling, converting, visualising, creating, and describing. The implications of the findings with regard to statistical enquiry pathways are discussed.

There is a need to explicate the design of learning tasks that introduce coding to introductory-level data science students. Computational lab components of an introductory statistics and data science course at the University of Auckland were written so students could complete them online. The R package learnr was used to create interactive lab tasks that featured videos, code exercises, progressive revealing of task components and quiz questions. This paper provides teachers with practical guidance on how to design and implement "first experience" tasks with coding using learnr. Consideration is given to balancing the learning of new statistical, computational, data-related, and tool-related knowledge.

The explosion in availability and variety of data requires learning experiences that reveal more of the data world faster and develop practical skills with digital technologies. Key high‐level goals of the International Data Science in Schools Project (IDSSP) include having students continually immersed in the cycle of learning from data, and data science being fun to teach and fun to learn. We advocate curiosity‐driven, exploratory learning for pursuing these goals. Our illustrations use tasks embedded in contexts that teenagers can relate to, provide visual rewards for computational actions, use rich data‐contexts, and integrate statistical and computational thinking. They provide engaging introductions to modern data sourced from databases via Application Programming Interfaces (APIs) that are accessible to a broad range of students and facilitate student personalization for investigation. We provide in‐depth discussion of teaching strategies that heavily involve questioning and student tinkering supported by graphical‐user interfaces that enable students to interact with the data sources rapidly in multiple ways.

The advent of data science has led to statistics education researchers re-thinking and expanding their ideas about tools for teaching statistical modeling, such as the use of code-driven tools at the secondary school level. Methods for statistical inference, such as the randomization test, are typically taught within secondary school classrooms using GUI-driven tools. A teaching experiment was conducted, using a learning task designed to introduce teachers familiar with using GUI-driven tools for teaching the randomization test to code-driven tools. Our findings from this exploratory study indicate that the design principles and considerations used to create the learning task supported teachers’ introduction to code-driven tools and encouraged an integration of statistical and computational thinking.

Informally testing the fit of a probability distribution model is educationally a desirable precursor to formal methods for senior secondary school students. Limited research on how to teach such an informal approach, lack of statistically sound criteria to enable drawing of conclusions, as well as New Zealand assessment requirements led to this study. Focusing on the Poisson distribution, the criteria used by ten Grade 12 teachers for informally testing the fit of a probability distribution model was investigated using an online task-based interview procedure. It was found that criteria currently used by the teachers were unreliable as they could not correctly assess model fit, in particular, sample size was not taken into account. The teachers then used an interactive goodness of fit simulation-based visual inference tool (GFVIT) developed by the first author to determine if the teachers developed any new understandings about goodness of fit. After using GFVIT teachers reported a deeper understanding of model fit and that the tool had allowed them to take into account sample size when testing the fit of the probability distribution model through the visualization of expected distributional shape variation. Hence, a new informal test for the fit of a probability distribution is proposed.