Learning R from R-Instat?

[rdstern]

Our proposition is that new courses on data science and statistics that use R could benefit from starting with a GUI. There are currently 9 GUIs in the set reviewed by Bob Muenchen.

R-Instat is aiming to be(come) sufficiently easy to use that it appeals to a wide audience in training courses. It may not be the easiest, but it is easy enough. One test of the ease of use could be through the Maths Camps, where I expect R-Instat to be used in 2023. These are teachers and secondary school-children, some of whom have little - or no - prior computing experience.

We usually assume initial computer literacy, so ease of using a mouse and perhaps some familiarity with a spreadsheet. Then it seems easy to add R-Instat. But the review in early 2022 emphasised that some might start with R-Instat, and perhaps by wishing to enter (a small amount of) data directly into R-Instat.

We then make the case, via Tutorial/discussion 3, that introduces the R-Instat calculator, that even that may not be too difficult, despite being a halfway dialogue. But it can be avoided for quite a time later, by those who find it difficult.

We have various halfway dialogues in R-Instat, with their narrow definition being that you enter a command - so you can learn about R-commands through this.

Now I want to take this idea further, and towards the possible courses on learning of R once you are already comfortable with R-Instat.

The next step after being able to understand about simple individual commands is that some dialogues may actually produce a script - or a very complicated command that is a sort of script in itself.

And scripts can be turned into functions. And functions can then add documentation and go into a package. And that is the strength of R, namely that there are many packages you can use.

And a great thing about open source is that you can even see the source code. So you could read the R-code that is in many functions in the packages.

An example is perfect numbers, that is easy up to the 10th perfect number, but takes forever, if you were to ask for 11. I wonder why? Looking at the function it is obvious. (Show in output window.)

Now suppose you would like to be able to calculate up to 12 perfect numbers. With his method the changes needed are obvious and simple. The encyclopedia lists them up to 15!

Below I step back and consider the proposed course more generally. I also have another discussion topic on the functions section, namely #8034.

But after functions comes packages and the above example could figure in that section. It is useful that you understand what a package is, and also how you can even see the code for a function in a package. And you could tweak that function if you wanted to improve it, as mentioned above - then use it instead of the one supplied.

But that is after being able to learn about the functions in an existing package to be able to use it well - with our halfway dialogues. And this can be extended so you can add a new package and still use those functions with our halfway dialogues, or in a script.

Then you can investigate a function, as mentioned above and even consider adapting it if you wish.

[rdstern]

Now I am trying to put this all together. This course is particularly for those who may still - at least sometimes - continue to use R-Instat. It is designed more for Middle and even Yuk, rather than for Love.
Most R courses are designed for Love and she may already have learned to use R, or Python herself. If not, then she may prefer to learn R (or python) from one of the books or courses that teach it, from the beginning. One example is the Stats4SD course here. This is given regularly - at least as we write these notes - and the materials are freely available (are they?).

You have already used some of the halfway dialogues, where you provide a command - and we review and extend there use later. So we start here with scripts.

Scripts:
There are 3 ways you can get a script. a) You can import it. b) You can complete a dialogue and then press the script button. c) You can write it yourself.

I suggest that might be our order? Though perhaps writing a simple script might also work, and then discussing what it doesn't do?

For importing I like the example from the Jane Austen books of a neat script. But it is quite complicated. But this is what the standard books do. So could be more useful examples where we then add bits so the results are in a data book.

Once we have finished the section on scripts, we could continue to functions and packages. Or we could have a review section on what readers should already know - from R-Instat - like factors and characters and numeric columns and data frames, etc.

In all this we want to emphasises what Yuk could usefully know at the end of it all. But it is particularly targeted for Middle. Let's be clear, with an example. Suppose Middle and Yuk are both doing their MSc and that includes a research project. They are using R and scripts that are used in their work are provided, with explanations, in an appendix.

Much of the work may have been done interactively, but the scripts, perhaps even the log file, are also provided.

Middle is keen to do the work herself. She doesn't see herself as an R programmer later, but this is to be her work, and she is keen to do it largely independently.

Yuk recognises that some R scripts could be useful, but needs help, because he really finds this part to be difficult. So he explains what he wants to Love, who writes most of the scripts for him. However, Yuk understands what Love has done sufficiently to write the explanations. In his report, he also acknowledges the role Love has played. The MSc is about the data and the problems solved, and not intended as a programming task. So he gets a good mark. So, from this course, he does need enough R skills to know when the dialogue approach is insufficient, and then to specify what is needed. Hopefully he knows enough that when he gets help, he is clear whether he is asking for a 1 hour, or one week task. If not this help might be needlessly expensive.