ADI has grown from reflections on past experience (Stern D. , 2014) which proposed the opportunity of ‘Viral Scaling’ as a route to sustainable impact. When working with potentially dynamic institutions a key point is that achieving sustainability within one can be harder that scaling across multiple similar institutions. So scaling can be a mechanism to increase the sustainability as well as the extent of the impact. For viral scaling the innovation needs to be able to spread independently, being passed from one user directly to another. An initiative can only go viral if an average user instigates more than one new user. Keeping this in mind has pushed R-Instat towards many of the other principles. It needs to be free to distribute (Open by Default), to be thought of as part of a set of resources appropriate for multiple audiences (Options by Context) which can stand alone and be used without needing specific training (Making it Easy).

The principle of Open by Default is that no justification is needed to go open. Not going open is also fine as long as there is a justifiable reason.  The decision for R-Instat to be open source was easy and obvious. Building from R we are “standing on the shoulders of giants” and the intention is for R-Instat to enable more people to make use of R. Harder decisions concerned choices of development tools and which platform to develop for. There was a strong desire to build from existing open source R interfaces, particularly R-Studio, which would have led to a cross platform solution. However, the aim of having the development done substantially in Africa was important. Development in VB.Net using Visual Studio has enabled us to produce a stable product while building the team. Other alternatives had a learning curve that was too steep. R-Instat is not intended as an open replacement for the standard commercial statistics packages. In resource rich environments, students should ideally become familiar with more than one package and R-Instat might sometimes be a useful addition to the mix. In resource poor environments R-Instat combined with R-Studio is designed to be sufficient for student needs.

R-Instat is designed to be as easy as possible for users to engage with.  It should be accessible for anyone who has experience with the use of a spreadsheet package.  This aim has driven our plan to make the individual R-Instat dialogs easy to use.  When more flexibility is needed this is usually through a sub-dialog. Developing the graphics system for R-Instat has been particularly time-consuming.  We believe strongly in the grammar of graphics (Wilkinson, 2005) and in its implementation in R, ggplot2 (Wickham, 2010).  However, even some experienced R users find ggplot2 difficult to master and shy away from it. Our challenge was to make the graphics dialogs easy and intuitive while also giving access to the immense flexibility of the ggplot2 system. A second aspect is to make it as easy as possible for educators to use interesting data in their teaching.  Having access to a rich library of datasets in R-Instat is critical and is helped considerably by the rich variety of datasets provided in R-packages.

The idea and phrase “Options by Context” has come from research methods work related to agricultural research (SSC, 2015). The phrase has influenced the look and feel of the software substantially as we recognize that not all target audiences will use R-Instat in the same way.  This has, in particular, influenced the decision to include tailored menus for specific audiences. Options by context also applies to our ideas for educational initiatives. This leads to an expectation for educational resources to be used in different ways, at different institutions, by different types of learners. This combines well with the open by default principle and implies that resources created should be editable where possible, so they can be adapted by users to their own contexts.

The idea that educational reform can be more effective when considered across all academic levels resulted from work in post graduate education which led to opportunities for that reform to spread to other levels (Stern D. , 2013). From basic statistical literacy at school to PhD students, the aim, partly through ADI, is to spark reforms which lead to substantial impact on the understanding and use of data. The first advantage from this approach is because many of the people trained in African postgraduate education are involved in education at undergraduate or school-level. Hence if the reforms they experience can translate into the classes they teach they are more likely to become part of the reform process.  The second advantage is that reform across academic levels encourages the cross-fertilization of materials and pedagogies. This is particularly important in making the teaching of basic material more exciting and interesting.

In many universities statistics courses need to change fundamentally to produce genuinely useful graduates.  However, wholesale changes are often slow and contentious to implement.  There is a risk that attempting large changes might meet with sufficient resistance that nothing changes. In contrast, small changes, starting within a single course, can be useful as well as being a precursor to more substantive change.  These small changes are often possible within the current syllabus, so changes in the curriculum may follow initial changes in the content or in the style of teaching.  Some could even be started within a mathematics or statistics “club”, as an extra-curricular activity, rather than in a formal course. The process of change through incremental improvements is potentially a cultural shift within African universities which could be a ‘win-win’ for both staff and students. If staff find ways to improve student educational outcomes, document their innovations and publish the results, then students have a better education, while the staff can use the resulting publications for their reputation and promotion. This could lead to a cultural shift whereby educational innovation is institutionally supported and valued.

At the heart of ADI is the desire to promote and encourage good statistical practice. Within R-Instat, decisions ranging from the menu structure to the choice of R packages are designed with this principle in mind. Although there is agreement on aspects of good statistical practice, such as the inclusion of descriptive comments in scripts or moving beyond averages (Ross, 2016), it is not always well defined and even within the development team this had led to rich and interesting discussions. The launch of R-Instat is designed to open up the discussion on what constitutes good statistical practice in different contexts and how this might translate into the software. R is already excellent at encouraging such rich discussions and the intention with R-Instat is to conclude the discussions with an implementation of the ‘good current practice’.