Pit templates #12
Pit templates #12
Conversation
| @@ -167,6 +167,6 @@ def obs_data(): | |||
| "location": ["a"] * 20 + ["b"] * 20, | |||
| "population": [100.0] * 10 + [150.0] * 10 + [200.0] * 10 + [250.0] * 10, | |||
| "age_group": (["young"] * 10 + ["old"] * 10) * 2, | |||
| "date": [datetime.strptime("2020-01-01", "%Y-%m-%d") + timedelta(i) for i in range(10)] * 4, | |||
| "date": [datetime.strptime("2020-01-14", "%Y-%m-%d") + timedelta(i) for i in range(10)] * 4, | |||
There was a problem hiding this comment.
needed to update date here to get observed data that were in the same range of prediction target dates, so that I could test correct results based on PIT scores.
| reference_time_col: str = "reference_date", | ||
| horizon_col: str = "horizon", pred_col: str = "value", | ||
| idx_col: str = "output_type_id", |
There was a problem hiding this comment.
these have moved to the .fit method instead of .transform
| horizon_col, idx_col, pred_col) | ||
| min_horizon = model_out[horizon_col].min() | ||
| max_horizon = model_out[horizon_col].max() | ||
| wide_model_out = self._pivot_horizon(model_out) |
There was a problem hiding this comment.
reference_time_col, horizon_col, idx_col, and pred_col are now saved as attributes of the object, so we no longer need to pass them around as arguments and we use self.horizon_col, etc.
| if self.model_out_train is not None: | ||
| wide_model_out_train = self._pivot_horizon(self.model_out_train) | ||
| else: | ||
| wide_model_out_train = None |
There was a problem hiding this comment.
refactoring to-do: move this inside of self._build_train_X_Y. see #2
| time_col: name of column in `df` that contains the time index. | ||
| obs_col: name of column in `df` that contains observed values. | ||
| feat_cols: names of columns in `df` with features | ||
| target_data_train: pl.DataFrame |
There was a problem hiding this comment.
Checked the docstring types against the types in the function parameter 👍
|
|
||
|
|
||
| def test_build_train_X_Y_pit_templates(obs_data, wide_model_out, monkeypatch): | ||
| # we use monkeypatch to remove abstract methods from the |
There was a problem hiding this comment.
This is a neat trick! At some point, it might be worth creating a fixture for this pattern, since it's in the code base quite a few times.
There was a problem hiding this comment.
good idea, i filed issue #17 for that
This PR makes it so that templates for the Shaake shuffle can be constructed based on PIT values from past forecasts rather than just past observed data. This is also an intermediate step toward estimating copulas.