Deprecate describe_across in favour of skim, table1, gtsummary etc

[mark-andrews]

The describe_across function was an attempt to provide a summary statistics table. It does this but other packages provide much more functionality.

For example, the skimr::skimr does all of this and more.

And to produce formatted summary tables, the packages::commands

• table1::table1() (see here)
• gtsummary::tbl_summary (see here)

will do things far beyond anything describe_across can do.

It is possible, of course, that we wrap some of the above mentioned commands, or maybe re-export them.

[jensroes]

Just to say, we introduced in year 1 describe_across yesterday and some students actually said that this is so much better than doing everything individually like in describe.

I agree that there are other good options. The problem that I see with functions like skimr is that it gives too much information and you know that there is a tendency (also among researchers, certainly the ones from the SPSS generation) do just report everything their software gave them without thinking about what they actually need and how numbers related to what they do. I like about what we are currently doing that it is very obvious which students just copy code, run it and paste results in their report and which students think about what they do.

Having said that, showing students that outputs can be formatted nice in R would be really useful.

[mark-andrews]

I find describe_across useful, but everything it does can be done with other tools and those other tools do more too. You are right that some students will use those tools blindly and just copy and paste generic output. However, tools like skim can be customized to include some things and include other things.

As a general point, there are a large set of related tools that we should think about. Some tools, like skim are for quick interactive work, kind of like `glimpse. Other tools are for producing well formatted tables as they appear in reports. I think we can provide more tools in psyntur for doing either or both of these, and in addition or alternatively, we can recommend that other tools in other packages be used instead when doing these things.

The exact same point can be made for model summaries. The summary function can be improved. The broom packages provides better tools. We can wrap some of these tools in psyntur, both for doing interactive work and for producing well-formatted tables in reports.