issue:fix for issue #690 - incorrect computation of n

[StatKumar]

This is the fix for the issue # 690.

Math error was occurring when the len(n) = 1. While slicing the array, the last value of the array is not considered, and due to slicing, a variable's value was set to 0, and later in the code, log of that variable was taken, which led to math error.

Fix:
n = np.sum(~missing[firstnonmiss:lastnonmiss])
changed to
n = np.sum(~missing[firstnonmiss:])

[CLAassistant]

All committers have signed the CLA.

[jmoralez]

Hey @StatKumar. Can you please:

• Read the contributing guidelines and make the change in the notebook.
• Make the change I suggested here (lastnonmiss = missing_idxs.max() + 1). The proposed change here would unnecessarily break the code for people who are directly using the AutoARIMA model and passing arrays like [nan, 1, 2, 3, nan].
• Also change the other places where this is happening, which I referenced here as well.

[StatKumar]

@jmoralez Thank you for the guidelines. Will update soon.

[jmoralez]

Can you update this PR instead of opening a new one?

[StatKumar]

Sure @jmoralez

[review-notebook-app]

Check out this pull request on 

See visual diffs & provide feedback on Jupyter Notebooks.

---

Powered by ReviewNB

[StatKumar]

@jmoralez Updated the pull request with a new commit. Please let me know if anything else needs to be done. Thank you for taking time in guiding me through my first issue in statsforecast.

[jmoralez]

Thanks @StatKumar! This also happens in the myarima function. Can you add the fix there as well? Also please add a cell at the bottom with the original example from #690 (AutoARIMA().fit(np.array([1.0]))) to make sure that the fix works.