ndphillips / FFTrees

An R package to create and visualise fast-and-frugal decision trees (FFTs)
https://journal.sjdm.org/17/17217/jdm17217.pdf
135 stars 23 forks source link

Add additional sub-plotting options #25

Open ndphillips opened 7 years ago

ndphillips commented 7 years ago

in plot.FFTrees(), add options to only show ROC curve, and main tree with icons

ndphillips commented 7 years ago

Need to restructure the layout of plot.FFTrees pretty heavily. It's a bit of a mess right now :)

hneth commented 2 years ago

Started to re-structure plot.FFTrees(), but splitting into multiple parts (e.g., ROC curve as separate function) may facilitate maintenance.

When trying to call showcues() with data = "test" (e.g., plot(heart.fft, data = "test", what = "cues") for heart.fft with test data) I only get the Error: "There are no test statistics for this object."
Is this a bug or a feature?

hneth commented 2 years ago

The release of FFTrees 1.7.0 has fixed the most obvious plotting issues, but the corresponding code is still pretty bulky and should be simplified further. Beyond additional cleanups, I'd advocate using more intuitive plotting options.

Issue: The interplay between what and stats seems redundant, as the only function of stats is to hide some options. Also, the default of what = "tree" is counter-intuitive, as the real default appears to be to plot 'all' or 'everything', whereas the "tree" intuitively seems to denote the tree diagram in the center.

Suggestion:

This would leave the current defaults (i.e., showing everything, when nothing has been specified) intact, but allow for more nuanced selections, e.g., using what = 'tree' with additional options.

hneth commented 2 years ago

I'd consider the most immediate needs to be addressed in the recent release of FFTrees version 1.7.5 (on 2022-09-15).

However, as new ideas/needs may arise and plot.FFTrees() is still pretty clunky, I suggest keeping this issue open for now.