SaturationTransformation#

class pymc_marketing.mmm.components.saturation.SaturationTransformation(priors=None, prefix=None)[source]#

Subclass for all saturation transformations.

In order to use a custom saturation transformation, subclass and define:

function: function to take x to contributions
default_priors: default distributions for each parameter in function

By subclassing from this method, lift test integration will come for free!

Examples

Make a non-saturating saturation transformation

from pymc_marketing.mmm import SaturationTransformation

def infinite_returns(x, b):
    return b * x

class InfiniteReturns(SaturationTransformation):
    function = infinite_returns
    default_priors = {"b": {"dist": "HalfNormal", "kwargs": {"sigma": 1}}}

Make use of plotting capabilities to understand the transformation and its priors

import matplotlib.pyplot as plt
import numpy as np

saturation = InfiniteReturns()

rng = np.random.default_rng(0)

prior = saturation.sample_prior(random_seed=rng)
curve = saturation.sample_curve(prior)
saturation.plot_curve(curve, sample_kwargs={"rng": rng})
plt.show()

Methods

`SaturationTransformation.__init__`([priors, ...])
`SaturationTransformation.apply`(x[, dims])	Called within a model context.
`SaturationTransformation.plot_curve`(curve[, ...])	Plot curve HDI and samples.
`SaturationTransformation.plot_curve_hdi`(curve)	Plot the HDI of the curve.
`SaturationTransformation.plot_curve_samples`(curve)	Plot samples from the curve.
`SaturationTransformation.sample_curve`(parameters)	Sample the curve of the saturation transformation given parameters.
`SaturationTransformation.sample_prior`([coords])	Sample the priors for the transformation.
`SaturationTransformation.update_priors`(priors)	Helper to update the priors for a function after initialization.

Attributes

`model_config`	Mapping from variable name to prior for the model.
`prefix`
`variable_mapping`	Mapping from parameter name to variable name in the model.
`default_priors`
`function`
`lookup_name`