#990 - Plot FourierBase along date rather than index

[Ishaanjolly]

Enhancement: Plot FourierBase along date rather than index

Description

I added the following:

def sample_curve(
        self,
        parameters: az.InferenceData | xr.Dataset,
        use_dates: bool = False,
        start_date: datetime.datetime | None = None,
    ) -> xr.DataArray:
        """Create full period of the Fourier seasonality.

        Parameters
        ----------
        parameters : az.InferenceData | xr.Dataset
            Inference data or dataset containing the Fourier parameters.
            Can be posterior or prior.
        use_dates : bool, optional
            If True, use datetime coordinates for the x-axis. Defaults to False.
        start_date : datetime.datetime, optional
            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
        -------
        xr.DataArray
            Full period of the Fourier seasonality.

        """
        # Determine the full period
        full_period = np.arange(int(self.days_in_period) + 1)

        coords = {}
        if use_dates:
            if start_date is None:
                # Derive start_date based on the type of Fourier seasonality
                today = datetime.datetime.now()
                if isinstance(self, YearlyFourier):
                    start_date = datetime.datetime(year=today.year, month=1, day=1)
                elif isinstance(self, MonthlyFourier):
                    start_date = datetime.datetime(
                        year=today.year, month=today.month, day=1
                    )
                else:
                    raise ValueError("Unknown Fourier type for deriving start_date")

            # Create a date range
            date_range = pd.date_range(
                start=start_date,
                periods=int(self.days_in_period) + 1,
                freq="D",
            )
            coords["date"] = date_range.to_numpy()
            dayofyear = date_range.dayofyear.to_numpy()

        else:
            coords["day"] = full_period
            dayofyear = full_period

        # Include other coordinates from the parameters
        for key, values in parameters[self.variable_name].coords.items():
            if key in {"chain", "draw", self.prefix}:
                continue
            coords[key] = values.to_numpy()

        with pm.Model(coords=coords):
            name = f"{self.prefix}_trend"
            pm.Deterministic(
                name,
                self.apply(dayofyear=dayofyear),
                dims=tuple(coords.keys()),
            )

            return pm.sample_posterior_predictive(
                parameters,
                var_names=[name],
            ).posterior_predictive[name]

and following within each plot_* function as I was not able to define new attributes:

if "date" in curve.coords:
      x_coord_name = "date"
elif "day" in curve.coords:
    x_coord_name = "day"
else:
    raise ValueError("Curve must have either 'day' or 'date' as a coordinate")

Related Issue

• Closes Plot FourierBase along date rather than index #990
• Related to #

Checklist

• Checked that the pre-commit linting/style checks pass
• Included tests that prove the fix is effective or that the new feature works
• Added necessary documentation (docstrings and/or example notebooks)
• If you are a pro: each commit corresponds to a relevant logical change

Modules affected

• MMM
• CLV

Type of change

• New feature / enhancement
• Bug fix
• Documentation
• Maintenance
• Other (please specify):

---

📚 Documentation preview 📚: https://pymc-marketing--1068.org.readthedocs.build/en/1068/

[wd60622]

Thanks for taking this @Ishaanjolly !

Some initial comments and suggestions

Could you write a test for the sample plot workflow using dates

[wd60622]

Is it possible to not rely on the child classes? This makes an inheritance from the FourierBase not possible.

This can usually be solved by a method in each child class. For instance, get_default_start_date(today: datetime)

[codecov]

Codecov Report

Attention: Patch coverage is 55.26316% with 17 lines in your changes missing coverage. Please review.

Project coverage is 95.23%. Comparing base (d05c2d8) to head (c1be3a5).
Report is 2 commits behind head on main.

Files with missing lines	Patch %	Lines
pymc_marketing/mmm/fourier.py	55.26%	17 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #1068      +/-   ##
==========================================
- Coverage   95.85%   95.23%   -0.62%     
==========================================
  Files          39       39              
  Lines        3934     3969      +35     
==========================================
+ Hits         3771     3780       +9     
- Misses        163      189      +26     

Flag	Coverage Δ
	95.23% <55.26%> (-0.62%)	⬇️

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[wd60622]

Subtlety here is that user can pass date that doesn't start at beginning of month or year and the plot starts.

I think that's fine if someone adjusts the zero point in x / dayofyear. Though conventionally that'd be beginning of month or year.

Just thinking out loud

[wd60622]

Looks good! Thanks @Ishaanjolly