python - Indexing a pandas.Panel counterintuitive or a bug? -


I am having trouble understanding the indexing of a pandal. Panel (3D Data Structure) It says that works as indexing:

Obtaining value from a multi-axis object that uses the following notation (for example Using .loc on, but applies to .iloc and Ix too) Any of the axes can have an empty slice: Axis left out of the specification is considered:. (Egploc ['A'] is equivalent to PL ['A',:, ::])

p.loc [item_indexer, major_indexer, minor_indexer]

P> Now I will assume that the order of the remaining index is not changed when the dataframe is drawn, but:

(2, 3, 4) PILOok [0]. Size # Basic order out [5]: (3, 4) p. LOOM [2]:, 0] .shape # transposed out [6]: (4, 2) p. Loc [:,:, 0] .shape # transposed out [7]: (3, 2) p. Iloc [,, 0, [8]: (4, 2) p. Loc [1:, 0,:]. Size # Slicing item_inxter, then move out [9]: (4, 1) P.I.Looc [1:, 0] Expecting to get .shape # in the same form [9], but instead of slicing minor_enxor Line Out [10]: (3, 2)

Any idea why dataframe is changed while indexing main_index or short_index, but not item_index? And why is it different from the last example?

There has been some discussion recently. It seems that pandals are not very well prepared in the direction of en-dimensional data (N> = 3). An optional module, which is like pandals but otherwise for ND data, you can use plain pandes (2D) dataframe with Multi Index to simulate ND data.


Comments

Popular posts from this blog

apache - 504 Gateway Time-out The server didn't respond in time. How to fix it? -

c# - .net WebSocket: CloseOutputAsync vs CloseAsync -

c++ - How to properly scale qgroupbox title with stylesheet for high resolution display? -