cat#

superglm.cat( column: str, *, base: str = 'most_exposed', grouping: LevelGrouping | None = None, levels: Any = None, unseen: Literal['error', 'base'] = 'error', ) → BoundTerm#

Declare a categorical effect with a reference level.

Use cat("region") for categories, including categories stored as numbers. A bare string in a predictor always declares a numeric linear term and does not infer categorical encoding from the column’s dtype.

Parameters:

columnstr: Name of the categorical input column.
basestr, default=”most_exposed”: Reference level. Use the level with the greatest total sample weight, "first" for the first level, or a specific level name.
groupingLevelGrouping, optional: Combine input levels into groups before encoding.
levelssequence, data column or categorical dtype, optional: Declare the allowed input levels. With grouping, these are the original levels before grouping.
unseen{“error”, “base”}, default=”error”: Prediction policy for levels outside the fitted level universe. "base" uses the reference level and emits a warning.

Returns:

BoundTerm: A categorical declaration for the named column.