FixedScaling

pydantic model pymc_marketing.mmm.scaling.FixedScaling

Use a user-supplied constant that stays the same across model refreshes.

Examples

Fixed scalar scaling for production stability:

FixedScaling(dims=(), value=10_000.0)

Per-dimension fixed scaling (multidimensional MMM):

FixedScaling(
    dims=("country",),
    value={"US": 50_000, "UK": 30_000},
)

Multi-dimensional fixed scale (e.g. country × channel) with xarray:

import xarray as xr

FixedScaling(
    dims=(),
    value=xr.DataArray(
        [[1e3, 2e3], [3e3, 4e3]],
        dims=("country", "channel"),
        coords={"country": ["US", "UK"], "channel": ["tv", "search"]},
    ),
)

Long-format table via from_long_dataframe():

FixedScaling.from_long_dataframe(
    dims=(),
    df=long_df,
    value_col="scale",
    dim_cols=["country", "channel"],
)

Methods

FixedScaling.__init__(**data)	Create a new model by parsing and validating input data from keyword arguments.
FixedScaling.from_dict(data)	Deserialize; restores encoded xarray.DataArray values.
FixedScaling.from_long_dataframe(dims, df, ...)	Build fixed scaling from a long table (one row per coordinate combination).
FixedScaling.scaling_description()	Human-readable summary of the scaling strategy.
FixedScaling.to_dict([_orig])	Serialize to a dict via Pydantic model_dump.

field dims: str | tuple[str, ...] [Required]

The dimensions to perform the operation through ("date" is always included implicitly).

field value: float | dict[str, float] | xr.DataArray [Required]

Fixed scaling constant(s). A single float applies uniformly. A dict maps coordinate labels along the single remaining dimension after reducing over date and dims (see the multidimensional MMM). If more than one non-reduced dimension remains, use an xarray.DataArray whose dimensions broadcast to that grid (e.g. a vector over country when the media grid is country × channel). All values must be positive; NaNs are not allowed.