Capacity building in R when starting with R-Instat

[rdstern]

@anastasia-mbithe, @berylwaswa , @lilyclements and @rachelkg I'm starting this discussion partly because the Tricot participants were clearly impressed with the ease that R scripts can be used for analysis when using R-Instat. Beryl is currently working on some of the R scripts provided in the vignettes of the R packages.

There was also discussion, in the workshop, that they should become 5 days, rather than 3 days. So there will be opportunities to include this topic, at least for some participants, within the workshop. How should we do this?

And we have a similar situation with our climatic work. There R-Instat is accepted, but there is considerable scope , where there is scope for offering advanced capacity-building nand this could well include how to use scripts well.

We have already started with resources in this area. But I would like to suggest we are moving nicely towards a coherent way of providing these general training materials and also providing special materials for given target audiences.

In the introductory materials we are working towards the four tutorials, and they are now general (and comprehensive) background materials, for training workshops. In the Tricot workshop we used their own data from the start. The "standard materials" were mentioned, and shown only briefly. But they are mainly available as general supporting materials for those who feel they need a more solid foundation in R-Instat, to use it easily.

I suggest we work towards a parallel approach when introducing the use of scripts with R-Instat. We have already started with the general materials. I also have an interesting example of (maybe) a general - slightly more advanced - point that we have yet to discuss. That's the main reason for this discussion topic now. But then the approach, in a workshop, is to use useful scripts with real (Tricot or Climatic) data from the start. For Tricot, these include the scripts Beryl is working on now, that are provided with the packages. They go into our Tricot-scripts directory.

Note we have never been against learning R and R scripts. What we claim is that inisting that everyone has to start with R and R scripts is daunting for some potential users. We are keen to start with data. Later, working efficiently and comprehensively with data will benefit from the use of neat R and useful R scripts for many potential users. For the real R-shy they will still benefit from understanding where their dialog-based approach is limiting them, and hence when they should look for help and support.

With all this in mind I found an interesting script point in the Likert graphs issue, #9330. It might become a possible topic for our general support materials. Here are aspects:

1. ggplot2 is currently the most downloaded R package - and by quite a lot. This supports our decision to concentrate on the ggplot2 graphics system.
2. Many developers have added functions and packages to make ggplot2 easier - at least for those who use them directly.
3. We have generally avoided those functions, because (like dialogs) they limit the users to those graph options within the function. So, we have generally decided to use "neat" ggplot2. Then it is always easy to add (tweak) the code to facilitate the full power of the grammar of graphics system. Our PICSA graphs are maybe our best example.
4. There are exceptions and our use of gglikert, from the ggstats package is one of them. This is relaively recent and we really like it - why:
   a) It seems well thought-out and comprehensive.
   b) It includes much of the flexibility and options of any ggplot2 graph.
   c) It produces a standard ggplot2 object. So we can use our main ggplot2 sub-dialog to add extra elements, such as a title, or colour scheme, or theme.
   d) In addition, it is a slightly odd graph in that it plots many y-variables at the same time. And it explains that "behind the scenes" it stacks the data first, and procudes a new data frame that fits within the usual ggplot scheme of a y and x! In our ordinary graph dialogs we allow this option - also slightly behind the scenes - if needed. That's the one y variable or many option.