37  Summary

If you take one thing from this book, take the shape of the workflow: construct, plot, decorate, save. An hv_*() constructor does the analysis and hands you a bare ggplot. You add scales, labels, and a theme. You save it for wherever it is going. The figure’s content is decided once and its appearance as many times as you have audiences, and because every recipe makes its own data, you can always go back to a working example and adapt it.

One thing is deliberately left for later: tables lean on gt directly until a companion hvtiRtables package gives them the same house treatment the plots get.

A version note: the random-forest chapters run on the ggRandomForests 3.x API (varPro priority, partial dependence, ROC), a line that is still moving — pin the package version when you reproduce them.

Treat the book as living. When you build a figure for a paper that is not here, or find a cleaner way to draw one that is, add it. The point of collecting these recipes was never to finish the list – it was to make the next figure easier than the last.