PyMC Wrapper

Wrapper to unify the generation of pymc marketing priors

source

HSGP

 HSGP (name:str, **kwargs)

Generalized prior interface for pymc-marketing. This generalizes the use of the apply method to any prior.

hspg = HSGP(
    "hspg",
    m=40,
    ls=1,
    eta=3,
    L=60,
    dims=("date",)
)

source

HSGPPeriodic

 HSGPPeriodic (name:str, **kwargs)

Generalized prior interface for pymc-marketing. This generalizes the use of the apply method to any prior.

source

YearlyFourier

 YearlyFourier (name:str, **kwargs)

Generalized prior interface for pymc-marketing. This generalizes the use of the apply method to any prior.

yearly_fourier = YearlyFourier(
    name="yearly_fourier",
    n_order=3,
    dims=("date",),
)

FourierBase.sample_prior

 FourierBase.sample_prior (coords:dict|None=None, **kwargs)

Sample the prior distributions.

Type Default Details
coords dict | None None Coordinates for the prior distribution, by default None
kwargs VAR_KEYWORD Additional keywords for sample_prior_predictive
Returns Dataset Prior distribution.

FourierBase.sample_curve

 FourierBase.sample_curve (parameters:arviz.data.inference_data.InferenceD
                           ata|xarray.core.dataset.Dataset,
                           use_dates:bool=False,
                           start_date:str|datetime.datetime|None=None)

Create full period of the Fourier seasonality.

Type Default Details
parameters arviz.data.inference_data.InferenceData | xarray.core.dataset.Dataset Inference data or dataset containing the Fourier parameters.
Can be posterior or prior.
use_dates bool False If True, use datetime coordinates for the x-axis. Defaults to False.
start_date str | datetime.datetime | None None Starting date for the Fourier curve. If not provided and use_dates is True,
it will be derived from the current year or month. Defaults to None.
Returns DataArray Full period of the Fourier seasonality. 
prior_sample = yearly_fourier.sample_prior(coords={"date": np.arange(0, 365)})
curve_sample = yearly_fourier.sample_curve(prior_sample)
curve_sample.sel(chain=0, draw=slice(10)).plot(x='day', hue='draw');
Sampling: [yearly_fourier_beta]
Sampling: []






















source






WeeklyFourier




 WeeklyFourier (name:str, **kwargs)




Generalized prior interface for pymc-marketing. This generalizes the use of the apply method to any prior.




source






Prior




 Prior (name:str, **kwargs)




Generalized prior interface for pymc-marketing. This generalizes the use of the apply method to any prior.




source






Data




 Data (name:str, dims:Optional[Tuple[str]]=None)




Generalized prior interface for pymc-marketing. This generalizes the use of the apply method to any prior.