DOC/BUG: correlation between series of different lengths

[matanmazor]

Hi,
Is it intentional that pandas does not through an error message when computing a correlation between two series of different lengths?
e.g. pd.Series([1,2,3]).corr(pd.Series([3,2]))
Thanks,
-Matan

[daskol]

It seems like a proper behaviour. Internally, pandas aligns serieses with .align() method and then uses numpy or scipy.stats in order to estimate correlation coefficients. Such behaviour is convinient if serieses have the same indexes.

I think that docs should be updated with note about series alignment.

[jackgoldsmith4]

pd.Series([1,2,3]).corr(pd.Series([1,2,3,55])) returns 1.0 because the 55 is removed after the inner join align(). This doesn't seem like correct behavior. Also, the underlying nancor method requires both objects to be the same length. Wouldn't it make more sense to tell the user that the objects should be the same length (via an error message) rather than altering the given objects before finding a correlation?

[jreback]

this is correct; all pandas operations align

this ticket is about updating the docs to make this even more obvious

[jackgoldsmith4]

take