ti#

superglm.ti( left: str, right: str, *, n_knots: tuple[int, int] | None = None, decompose: bool = False, ) → BoundInteraction#

Declare a tensor interaction between two smooth effects.

Both columns must have spline terms in the same predictor. For example, use s("age"), s("value"), ti("age", "value") together. The tensor describes their interaction; the two s terms provide the main effects.

Parameters:

left, rightstr: Names of two columns with declared spline terms.
n_knotstuple of int, optional: Interior-knot counts for the left and right tensor margins. By default, inherit the parent terms’ counts.
decomposebool, default=False: Separate the bilinear direction from the wiggly interaction so their penalties can be controlled separately.

Returns:

BoundInteraction: A declaration named "left:right" using the supplied column names.