He's ascribing a lot more "understanding" to chatgpt than it deserves. But I'm somewhat impressed with it explaining why @dontrun is needed in the docs. #RStatsNYC#RStats
{ChannelAttribution} 📦 to get Markov chain attribution. Need a df of path + did they convert for each case, then it can do attribution cleanly and easily. #rstatsnyc#RStats
Considerations:
Do you have data (or can estimate)?
Markov chain assumes states are independent. If steps are dependent, consider modeling before and/or after dependent steps.
When does a customer journey restart? After purchase? Time based? https://github.com/JessicaMDuncan/2023_NYR #rstatsnyc#RStats
Friendship with {renv} has ended, {pak} is now bestfriend #rstats
My workflow for package management in R is now:
create a DESCRIPTION file (all projects are "packages")
use {usethis} to record new dependencies
use {pak} to create lockfile and install R + system dependencies
I use that inside devcontainers, but the nice thing is I am not tied to anything. Unlike a workflow that uses renv, which is required (and it had to do it's weird stuff messing up with .Rprofile)
2 weeks ago I had my windows reinstalled. It was pretty cool to just have to install git, vscode and docker to spin back a new devcontainer and get back to work.
Every project specific devcontainer depends on a custom base devcontainer image built and pushed on our gitlab image registry. Keeping each devcontainer json very minimal.
Once I am done with a project I can simply nuke the container, maybe keep the code locally, maybe not. Out of the disk, out of the mind.
Next steps will be to be able to sync shell history, style, custom recipes and other stuff between my host and every devcontainers I use (many, at least 1 per data analysis project) in a way that my colleagues can do the same thing if they want and with their own config.
@paezha@benleis@phonner@meghansharris@jacquietran@georgemsavva Yeah that was me, in a twitter thread in the days of yore. The term was originally intended to have wider applicability than generative art: my original conception was that it should refer to any use of #rstats for artistic or craft purposes (including art about R), but was mostly taken up by generative artists, so now it seems have acquired a narrower meaning where, in practice, it refers mostly to generative art made in R.
#rstats hivemind - what's the best way to save multiple objects from the global environment (i.e, tibbles) to .rds? So far all I've managed to save to the .rds files are the names of the tibbles, sans content.
@john you can use varhandle::save.var to save variables from an environment in independent RDS files. Disclaimer: it is my own #RStats package and have been around since R version 2.x. It is a very light package with no external dependency and straightforward functions:
@krzysztof_dyba@eliocamp@rstats That's an interesting & long review, pointing out lots of weaknesses in R's documentation & implementation. E.g., "from the solutions to Advanced R:
“Note that as.vector() and is.vector() use different definitions of ”vector!””.
The above quote is then followed by showing that is.vector(as.vector(mtcars)) returns FALSE."
Where one could open pull requests to help improve the base documentation?
Are you an R programmer looking for a powerful tool to compare objects for exact equality? Look no further! Say hello to the incredible identical() function in R. 🤝🔍
In my latest blog post, I dive into the world of identical() and uncover its syntax and superpowers in a simple and engaging way. The function allows you to determine if two objects are exactly identical.
Jessica Duncan's "Marketing Channel Attribution" talk is superb and is def going to help my team work more efficiently with Marketing folks at GN. #nyr#rstats
Amazing time hanging with the NY R #rstats conference speakers last night! Those incredibly talented humans are going to crush it today (y'all shld 100% have FOMO). We'll see if yours truly can not embarras them.
Link to live and source slides for my "Into the WebR-Verse" talk shld auto-post around 14:00 ET, here.
Def 👀 them since there is plenty of room for everyone in the WebR-Verse!