Asymmetric error bars are not supported for series (only dataframes) (original) (raw)

According to the documentation, asymmetric error bars should be able to be passed in to plot as an Mx2 array for a Series object or a Mx2xN array for a DataFrame object.

However, the code only seems to parse this correctly for a DataFrame: https://github.com/pydata/pandas/blob/master/pandas/tools/plotting.py#L1338

When passing raw error values as yerr, it seems that the first values are used as two-sided error differences, instead of being plotted as raw upper and lower values. Note the symmetric error bars below:

Using the underlying matplotlib function directly yields the expected results, but without the pandas-generated decorations:

This was originally noted in this StackOverflow question and I corroborated the other user's experience.