Infer the Stevens NOIR measurement level for a single vector.

Decision rules, in order:

1. Character / factor with n_unique <= ordinal_threshold and an ordinal name hit (likert/grade/scale/...): "ordinal".

2. Character / factor otherwise: "nominal".

3. Logical: "nominal".

4. Numeric with n_unique <= 2 (binary): "nominal".

5. Numeric with n_unique <= 20 + ordinal name hit: "ordinal".

6. Double with interval name hit (year/index/date/...): "interval".

7. Double otherwise: "ratio".

8. Integer with non-negative range: "ratio"; else "interval".

9. Date / POSIXct: "interval".

Usage

morie_dataset_infer_level(x, name = NULL, ordinal_threshold = 10L)

Arguments

x

A vector (any atomic type or factor).

name

Optional column name to drive the name-based heuristics. Defaults to NULL (no name-based promotion).

ordinal_threshold

Integer; max unique values for a categorical column to be considered ordinal (default 10).

Value

Character scalar; one of "nominal", "ordinal", "interval", "ratio".