[Feature Request]: Make faceting optionally in a grid rather than just one column in tm_g_km

[danielinteractive]

Currently the faceting in tm_g_km only works with one column. The plots get very "flat" then.
Sometimes it might be better to use a grid e.g. with 2 or 3 columns.

[danielinteractive]

@shajoezhu @lcd2yyz fyi

[mthbretscher]

The core issue is that "faceting" produces very disfigured (hight vs width) plots - they become very flat - and there's no one-size-fits-all setting for plot dimensions that could mitigate that. Either arrange faceting in a grid, or dynamically increase the plot height (multiply by the number of facets) could help..

[mthbretscher]

I found out a bit more on this - facetting actually works when a factor has only few levels (e.g. gender, with two levels). If I try to facet by a variable with many more levels (e.g. treatment category), then the plot height appears to be determined by total number of levels in that factor variable - not by the actual number of levels that are present in the data. E.g. if I filter to retain only 2 treatment categories - two facet plots should still look great. But it's doing like 10 plots (because the variable has this many levels).. the resulting plots are so flat that nothing can be seen.. could one smuggle in a "drop_unused_levels" or similar within the KM model (after applying the filters, before plotting)

[mthbretscher]

One way I could guess how one could solve this:
at this line in the code, replace nrow = seq_along(levels(anl$facet_var)) with something that drops unused levels from anl$facet_var first.. (not tested)

[mthbretscher]

As a temporary fix for myself I chose to override template_g_km template_g_km . This appears to be working reasonably well for a start (i.e. one can facet plot, just need to subset to a small number of categories at a time, and plots are still a bit flat). Faceting in a grid still worth considering.