Pandas：有条件地组合多索引数据帧

Question

I have two dataframes (df1 and df2) which are used pd.groupby from a same datafram. I want to conditionally create another dataframe (df_result) based on the following logic:

1. df_result's structure / cell is copied from df1
2. For count < 6 in df1, use the corresponding value in df2 instead.

INPUT1: df1

Type_lv1  Type_lv2          diff
                         count mean
t1           t_a              1    0.02
             t_b              12   0.01
t2           t_a              5    0.12
             t_b              22   0.11

INPUT2: df2

Type_lv1         diff
               count mean
t1             13    0.011
t2             27    0.11

OUTPUT:df_result

Type_lv1  Type_lv2          diff_result
                             mean
t1           t_a              0.011
             t_b              0.01
t2           t_a              0.11
             t_b              0.11

In sample data above, the first roq and third row in df_result come from df2 since the counts in df1 are 1 and 5, which are smaller than the threshold 6. I have tried to found related samples in past questions but could get what I want. Could someone give me some directions?

Thanks!!

Answer 1

Idea is add second level from df1 to df2 for MultiIndex , so possible repalce by condition in last step, only necessary matching first level of MultiIndex in both DataFrame s:

m = df1[('diff','count')] < 6

mux = pd.MultiIndex.from_product([df2.index.unique(), 
                                  df1.index.get_level_values(1).unique()])
df3 = df2.reindex(mux, level=0)

df1.loc[m, ('diff','mean')] = df3[('diff','mean')]
print (df1)
                   diff       
                  count   mean
Type_lv1 Type_lv2             
t1       t_a          1  0.011
         t_b         12  0.010
t2       t_a          5  0.110
         t_b         22  0.110