Conformal Prediction

[kvnkho]

Some things:

1. I am least confident about the MSTL implementation if you can take a took.

2. Am I right in thinking that prediction_intervals=ConformalIntervals(h=13, n_windows=2) means that you need to do forecast(h=13), otherwise the shapes of the intervals will not match and you can't form the output DataFrame. I feel like we should either catch this and raise an explicit error rather than a numpy shape error, or we can handle implicitly and override?

3. The type hint for level: Optional[Tuple[int]] causes a mypy issue because of the following behavior:

test = (1, 2)
print(type(test))            # tuple
print(type(sorted(test)))    # list

so the line:

levels = sorted(levels)

casts tuple to list, which mypy doesn't like. So I think all the levels hints should be Optional[List[int]]

[review-notebook-app]

Check out this pull request on 

See visual diffs & provide feedback on Jupyter Notebooks.

---

Powered by ReviewNB

[AzulGarza]

Thank you @kvnkho! This is awesome! 🥳

Could we clean the notebooks to merge the PR?🙌

[kvnkho]

Thank you @kvnkho! This is awesome! 🥳

Could we clean the notebooks to merge the PR?🙌

Finished cleaning I think.