ppc_ecdf_overlay(): geom = "step"

[fweber144]

In #234, we discussed the following:

• I tried to stick as close as possible to ppc_ecdf_overlay(), with one exception: In ppc_km_overlay(), I hard-coded the use of geom_step(). In contrast, ppc_ecdf_overlay() has an argument discrete which defaults to FALSE so that geom = "line" is used. I am using geom_step() in ppc_km_overlay() because in my opinion, the ECDF (and therefore also the ECCDF and the Kaplan-Meier estimate) is defined to be a step function. If you want, I can open a new issue for this as far as ppc_ecdf_overlay() is concerned.

I somehow lost sight of this. Have you thought about hard-coding the use of geom = "step" in ppc_ecdf_overlay() (and ppc_ecdf_overlay_grouped()) or setting the default of argument discrete to TRUE?

[utkarshpawade]

the ECDF is a step function by definition, regardless of whether the data is continuous or discrete. so geom_step() should be the default. ppc_km_overlay() already hard-codes step geometry for the same reason.
looking at the code, line 294 in ppc-distributions.R does geom = if (discrete) "step" else "line". the fix would be to always use "step" and deprecate the discrete argument with lifecycle::deprecate_warn() since it wouldn't serve a purpose anymore. ppc_ecdf_overlay_grouped() delegates to ppc_ecdf_overlay() so it gets the fix for free.