Feature/extra suggestions

[statzhero]

Outstanding work! I fully transitioned to R many years ago, but sometimes I'm still haunted by the ease of some Stata commands. Your "Rosetta Stone" already covers my main workhorses.

Here's a brain dump of features that I wish R (or I) could do better – with the full understanding that this is not a request to add them but more of a list to consult if you're looking to expand the page.

• Most R functions provide limited feedback. One solution is tidylog and in my opinion real-time feedback on data manipulation and merge operations is critical for integrity.
• Log files (there are some hacks in R but most of them are sad)
• Using weights in R is often cumbersome, especially more complicated types, or if you leave the core libraries.
• Panel data diagnostics, such as a test function whether the data is balanced or not.
• Binscatter plots and regression: binsreg
• Programmatically create variables, charts, or even just passing variable names into functions; something that works every time.
• Bootstrapping seems involved when used beyond trivial examples (see Stata bsample and Drechsler et al. 2021 for what I mean by non-trivial).
• And finally TABLES. (Even a simple, good-looking frequency or cross-tab can be a nightmare for me!)

p.s. tidytable for those who want to have the cake and eat it, although there's loss in many ways.

[grantmcdermott]

Thanks @statzhero. I think at least some of what you want is already in there. E.g. See the modelsummary and collapse suggestions in the "extras" page for tables. Some others I think are beyond scope despite obvious being useful (e.g. binsreg).

I'm not sure I quite follow the difficulties with programmatic variable creation or boostrapping. The former is easy to with base R tools IMO, but there's also fixest's macro facilities that we highlight at several points in the page. For the latter, is there something that sandwiich::vocBS or the boot package don't give? I do think data.table excels at bootstrapping simulations, but again that's maybe getting a bit to specific without a simple example.

Weight regression is a good idea. We should add that.

Log files I'm not keen on personally, since I don't think they make much sense for R (you can just recall / save different objects) or use Rmd for full literate programming. But we did talk at some point about adding a "miscellaneous" tips section where stuff like that could go in.

[statzhero]

Great – obviously my list is terse but if there are questions I can supply examples and use cases. But as I said, take what you like.

Re bootstrapping: it's my understanding that boot and other libraries don't integrate as well as bsample, e.g. when you use coefficient estimates in a second regression as variables or other multi-step simulations.

Re logs: I had a specific use case but more to the point, I was thinking of stuff they teach (or taught) in the Stata 101. Same idea with how to use the help files and vignettes.

One more question: I use modelsummary in a dubious cocktail of custom theme, knitr, and Latex hacks. Are you saying it's the best equivalent to tabout? (And I don't remember if or why it's different from estout.)