Improving the specific plots in R-Instat

[rdstern]

The dialogues in Describe > Specific provide one (perhaps the most used) way to produce good graphs (and tables) using R-Instat. In terms of their use, they are after the production of quick presentations using Describe > One Variable, Two Variables and Three variables. Those dialogues are designed to produce quick - usually exploratory graphs. They should help at the start of the analyses.

When you look in more detail, and may wish also to produce presentation graphs (and tables), then you use the dialogues in Describe > Specific.

To complete the context, you can go further. You can use Describe > General > Graph and build up the graph step (geom) by step. More generally still, of course the dialogues simply build a sequence of R-commands. You could start with commands from the script button on each dialogue, or just from the script and run the R-commands themselves. Of course, if you only want to run R commands, then we suggest RStudio instead. The nscript window is useful if you usually use R-Instat through dialogues , but occasionally, for convenience or efficiency, or reproducibility you choose to use scripts occasionally.

We have recently been guided by 4 books on data visualisation and data science. One aspect has been to be able to provide (almost) all the graph types from those books in the Describe > Specific dialogues. This is leading to the "specific" dialogues each offering a selection of related geoms (geometric shapes). These alternatives are sometimes via the top radio buttons, e.g. boxplot, jitter, violin. Alternatively they are through checkboxes,

a) with Histogram/density/frequency the density ridge-plots are a checkbox once you choose a density plot.
b) Tufte_boxplots are an optional checkbox once you choose boxplots.

We describe here, the continuation of this process. It is to be tested and improved during the forthcoming sprint. Some of the items below are already given as issues, while others are new and need discussion.

a) Dotplots - they become a checkbox on the histogram dialogue
b) Treemap and Mosaic are added become further radio buttons on the bar chart dialogue. Currently Mosaic is on its own menu item and treemaps are only on the procurement menu.
c) Lollipop charts are a checkbox on the bar charts - as are pie and donut!
d) At the same time we delete Pie from the name of the dialogue. It is only a checkbox now. We add the Treemap and Mosaic, to the name, so Bar/TreeMap/Mosaic Plot?
e) The options for different orders of the factors are bing added to the bar chart dialogue. This might be equally applicable to the other options there, and perhaps on some other dialogues that include factors.
f) I propose we also change the position of this dialogue. It is now important, and could well be "promoted" to be just under the line, i.e. before Point Plot.
g) Add a checkbox (user control for rug plot to the point plot. (Possibly later also to line plot. It may include some options, particularly to include only one axis. Then also omit rugplot from the Describe > Specific menu.
h) We have now included dumbbell as a geom, and there is a function for slope charts. So add them to line charts. Also, before that, add smooth. There will be an issue on this. So it will be Line/Smooth/Dumbbell/Slope.
i) One topic missing is heatmaps. This is an item where one of the books recommends the ordinary heatmaps, and they do look nice - and include dendrograms. In ggplot, geom_tile is used, so perhaps they could go together on a dialogue. They are mentioned together with geom_rect and geom_raster. Perhaps they could all go together?

In addition I suggest we add a control for facets to most, if not all the specific dialogues. We probably have this control elsewhere already, e.g. in the PICSA graph dialogue. It includes a pull-down with the default being wrap.

[rdstern]

In addition to the above topics I would like to use the sprint - and the lead-up to it, to improve specific aspects that have been around for a long time.

a) We need to be able to add equations information to our graphs. @dannyparsons already does this for some of the special climatic graphs, but it is not yet available in general. There seem to be 2 alternative packages that could provide the geoms and stats for this. One is ggpubr and the other is ggpmisc. I think both are in R-Instat. However, ggpmisc seems to have split into 2 packages from Version 4 with the other called ggpp. I thin k ggpp will automatically be installed when we update ggpmisc in the next version of R-Instat.

We tired with ggpubr in #5744 but that was never completed. I assumed ggpubr, but now think we should implement ggpmisc instead. That needs to be confirmed by @dannyparsons . The choice looks obvious to me - or maybe we use both. But I am slightly nervous, because the downloads of ggpubr are so much higher than those for ggpmisc.

b) We need to complete the work on the axes for different types of data. We have options for continuous data. For dates it is still not easy for seasonal plots - could this be an option that is added? And there is still nothing for discrete data.

c) I suggest there is more we need to add on scales. I think this is also the idea of colour palettes as proposed in #4806. I would like some circular palettes.