How Do I Change Or Access Pandas MultiIndex Column Headers?

I have the following Pandas DataFrame, but am having trouble updating a column header value, or easily accessing the header values (for example, for plotting a time at the (lon,lat

Solution 1:

You can use df.columns.get_level_values('lat') in order to get the index object. This returns a copy of the index, so you cannot extend this approach to modify the coordinates inplace.

However, you can access the levels directly and modify them inplace using this workaround.

import pandas as pd
import numpy as np

df = pd.DataFrame(columns = ["id0", "id1", "id2"])
df.loc[2012]= [24, 25, 26]
df.loc[2013]= [28, 28, 29]
df.loc[2014]= [30, 31, 32]

df.columns = pd.MultiIndex.from_arrays([df.columns, [66,67,68], [110,111,112]],
                                       names=['id','lat','lon'])

ids = df.columns.get_level_values('id')
id_ = 'id0'
column_position = np.where(ids.values == id_)

new_lat = 90
new_lon = 0

df.columns._levels[1].values[column_position] = new_lat
df.columns._levels[2].values[column_position] = new_lon

Solution 2:

You access MultiIndex via tuples. For example:

df.loc[:, ('id0', 66, 110)]

However, you may want to access via lon/lat without specifying id or maybe you'll have multiple ids. In that case, you can do 2 things.

First, use pd.IndexSlice which allows for useful MultiIndex slicing:

df.loc[:, pd.IndexSlice[:, 66, 110]]

Second:

df.stack(0).loc[:, (66, 110)].dropna().unstack()

Which is messier, but might be useful.

Finally, the last thing you mentioned. For a specific row with lon/lat.

df.loc[2014, pd.IndexSlice[:, 66, 110]]