Thread by @JonTheGeek

And now Aaron R. Williams will tell us about generating fake data (but for good) #RStudioConf2022 #RStudioConf #RStats

Crazy high unemployment spike on the cover of the nytimes. #RStudioConf2022 #RStudioConf #RStats

If we can make fake data that matches statistical properties of real data it can help with confidentiality. #RStudioConf2022 #RStudioConf #RStats

Ohhh neat. Start by sampling variable 1 with replacement. Make a model to predict var2 given var1. Then use 1 and 2 for 3. Etc. #RStudioConf2022 #RStudioConf #RStats

Can also set rules around how data should work, add additional noise, etc
#RStudioConf2022 #RStudioConf #RStats

One way to use it: tell it the most predictive bar, then have it add variables in order of how much they correlate with that variable. #RStudioConf2022 #RStudioConf #RStats

Can use recipes to set rules for processing the data. Has a default construct_recipes but you can expand. #RStudioConf2022 #RStudioConf #RStats

Gah I missed the photo of the pretty plots comparing real palmer penguins to a synthesized version. They're close but not identical, neat! #RStudioConf2022 #RStudioConf #RStats

He's @awunderground. Finding this on github, one moment... #RStudioConf2022 #RStudioConf #RStats

Doh, package isn't public yet, while he's careful about privacy. Later this year! #RStudioConf2022 #RStudioConf #RStats