It’s been said that there needs to be change in statistics education in schools. Do you agree?

Darren Macey: It has been suggested by many that the growth of computer science in industry and the growing role of data science in the commercial world is not currently well served by education, at least until you get to university level.

Alongside that is a recognition that data literacy is an absolutely fundamental part of engaging with society. There’s also a really interesting focus on non-standard representations of data (what most people know as infographics), which people are engaging with on social media.

I taught statistics for a long time, and my sense of what needs to change is about providing richer opportunities to ask more interesting questions and for students to ask questions themselves. For example, moving from current practice questions such as asking “What is the correlation?”, which for me is just a call and response situation, to asking: “What interesting correlations or patterns might you see here, and why?” or “Why does this patch of data look different?”

We want to model the idea that the interesting and important questions you ask in statistics are not about the data, they are about a problem for which you use statistical techniques on the data. It comes down to using statistics to answer questions, rather than answering questions about statistics, which is completely backwards.

Will Hornby: We support embedding statistics teaching within the statistical cycle, rather than just isolated techniques, bringing in explorative data analysis and the iterative nature of statistical literacy.

For example, instead of learning to calculate a whole load of measures of central tendency (mean, median, mode) as ends in themselves, instead you might start with a motivation which is grounded in context.

So rather than: “Calculate the mean of these heights”, you might ask: “How could I represent the height of this class with single number?” That can lead to discussion of what the measures tell you (or don’t tell you) about the data, not just how to calculate them.

As a teacher who wants to teach statistics for understanding, where can I start?

WH: I think a good starting point is to think about the technique you were going to teach and turning it around to consider how to provide a motivation for this tool. Can we ask some genuine and interesting questions that we can answer with data, rather than questions about calculations with data?

DM: I think there also needs to be a recognition that it’s not necessarily a linear process or something you tick off in a lesson. What you can do is think more across the term or year about how you are going to call attention regularly to bigger statistical ideas.

At the beginning, you may have to do lots of prompting, but by the end of the year students will be used to asking these sorts of questions because they have seen it modelled. They will know to consider patterns, shapes, clusters and lumps and bumps in the data.

Speaking of which, students should be encouraged to use their own descriptive language, which the teacher can then refine, because it allows them to freely articulate the instinctive understanding they will have about statistics. It is easy to get hung up on using formal, precise language all the time.

WH: The use of language is really important. I don’t mean always using formal language necessarily, but there are some basic things about inferential thinking that can happen through very subtle differences in the way you ask questions - for example, asking: what does this tell me about heights of students in this classroom?

DM: And teachers have to model that: “What can you say about…? How sure are you?”

What sort of technology best supports good statistics teaching?

WH: Technology has the really crucial benefit of easily moving between different representations (graphs, charts and diagrams) of the same data. This gives students an insight into flexibility and variability in a way that is impossible when drawing things out longhand. Technology allows for dynamic representations in order to examine the effects of change by manipulating representations.

You actually don’t need very much tech to completely transform the way you teach statistics, and anyone with the internet has access to a great deal of powerful tools. But a lot can be done very simply with pen and paper; it’s more about changing the approach, and technology is a great tool for that.

Do you have a favourite statistic or data set?

DM: The one I often use for professional development with teachers is the Mayfield High data, which is still available on the internet and is often familiar to students. It contains things like height, eye colour, favourite subject, so it’s a familiar context, rich in terms of different types of data, large enough to do interesting things with, and teachers can easily collect some data from students in their classes to compare.

WH: One of my favourites is the Gapminder Foundation stuff [Gapminder is a Swedish website that uses statistics to produce teaching resources that challenge misconceptions about global development], which is absolutely vast and truly exciting, with such a low threshold to entry. There are some amazing videos online of Hans Rosling charging around and being excited about the data.

DM: Yes, this is an excellent tool for telling stories from data. I’ve had teachers immediately texting other teachers in CPD sessions saying: “We need to do stuff with this right away!” because it really captured their imagination.

Will Hornby is an assessment standards manager for a major UK exam board and an associate lecturer for the Open University. Darren Macey works as a framework developer at Cambridge Mathematics. Their book, Teaching Statistics, is published by Cambridge University Press

This article originally appeared in the 12 July 2019 issue under the headline “How data science is going stats…to the future”