Handle Non-Daily Time Resolutions in _time_resolution Calculation

[paulogonc]

The current implementation of the _time_resolution calculation assumes that the data in self.X[self.date_column] is sampled at least in daily intervals. However, if the data is sampled at finer resolutions (e.g., seconds, minutes, or hours), the calculation self._time_resolution = (...).days results in a Zero.

if self.time_varying_intercept | self.time_varying_media:
            self._time_index = np.arange(0, X.shape[0])
            self._time_index_mid = X.shape[0] // 2
            self._time_resolution = (
                self.X[self.date_column].iloc[1] - self.X[self.date_column].iloc[0]
            ).days

When using time varaying intercept this will cause following error:

hsgp_kwargs.L = time_index_mid + DAYS_IN_YEAR / time_resolution
                                     ~~~~~~~~~~~~~^~~~~~~~~~~~~~~~~
ZeroDivisionError: float division by zero

Proposal to raise a ValueError after calculating self._time_resolution:

if self._time_resolution == 0:
      raise ValueError(
        "Time intervals in date_column are not in full days. Ensure that the dataset's timestamps are sampled at least in daily intervals."
    )

It took me a while to find out, that my timestamp data where sampled in the wrong resolution.

[wd60622]

What is the frequency of your data?

[paulogonc]

My data is typically sampled at a frequency of 1 second, with a duration ranging from 1 to 5 hours. To use TVD with this data, I resample it to a daily frequency.