将多索引列追加到空数据帧中

df = pd.DataFrame(columns=pd.MultiIndex(levels=[["foo","baa"],["x","y","z"]], labels=[[0,0,0,1,1,1],[0,1,2,0,1,2]]), index=[0,1,2,3], data=np.random.rand(4,6))

         foo                           baa                    
          x         y         z         x         y         z
0  0.263780  0.942880  0.176744  0.914854  0.920735  0.573692
1  0.601317  0.584691  0.288536  0.832064  0.095142  0.186045
2  0.807323  0.075620  0.399703  0.936894  0.168441  0.468984
3  0.871293  0.324817  0.395784  0.133541  0.365586  0.615219

r_df = pd.DataFrame(index=df.index)
Empty DataFrame
Columns: []
Index: [0, 1, 2, 3]

         foo
          x         y         z
0  0.263780  0.942880  0.176744
1  0.601317  0.584691  0.288536
2  0.807323  0.075620  0.399703
3  0.871293  0.324817  0.395784

我试着这样做:

for label in df.columns.levels[0]:
    data = df[label]
    for row in data.index:
        r_df.loc[row,(label,"x")] = data["x"]
        r_df.loc[row,(label,"y")] = data["y"]
        r_df.loc[row,(label,"z")] = data["z"]

我认为这可以工作,因为多索引可以被tubles引用,但事实并非如此。