Are pandas Panels as efficient as multi-indexed DataFrames?

Question

I am wondering whether there is any computational or storage disadvantage to using Panels instead of multi-indexed DataFrames in pandas.

Or are they the same behind the curtain?

Answer 1

they have a similiar storage mechanism, and only really differ in the indexing scheme. Performance wise they should be similar. There is more support (code-wise) for multi-level df's as they are more often used. In addition Panels have different silicing semantics, so dtype guarantees are different.