Noted this pattern while reviewing:

    out["med_schwierigkeit_treppen_pl"] = object_to_int_categorical(
        raw_data["ple0004"],
        renaming={"[3] Gar nicht": 0, "[2] Ein wenig": 1, "[1] Stark": 2},
        ordered=True,
    )

This is problematic because I'll always have to look up the definition when actually using the variable. Should read (I have not looked up whther this is valid code):

    out["med_schwierigkeit_treppen_pl"] = object_to_categorical(
        raw_data["ple0004"],
        renaming={"[3] Gar nicht": "Gar nicht", "[2] Ein wenig": "Ein wenig", "[1] Stark": "Stark"},
        ordered=True,
    )

Applies to all occurrences of this when the number of outcomes is small / all of them have meaningful names. In those cases, either we have an int, or a categorical.

Counterexample: Likert-scales, i.e., the question is literally "On a scale from 1 to 7 where 1 means ... and 7 means ..., what ..."

Edited after realising the Likert-scale case