You can also have multiple levels when reshaping your data. This will help you provide more specificity to how you want your data to be manipulated. If given the following data set:

exp                 A         B         A         B
animal            cat       cat       dog       dog
hair_length      long      long     short     short
0            1.075770 -0.109050  1.643563 -1.469388
1            0.357021 -0.674600 -1.776904 -0.968914
2           -1.294524  0.413738  0.276662 -0.472035
3           -0.013960 -0.362543 -0.006154 -0.923061

We can specify a compression of the levels animal and hair_length using df.stack(level=["animal", "hair_length"]). The resulting DF will be

exp                          A         B
  animal hair_length                    
0 cat    long         1.075770 -0.109050
  dog    short        1.643563 -1.469388
1 cat    long         0.357021 -0.674600
  dog    short       -1.776904 -0.968914
2 cat    long        -1.294524  0.413738
  dog    short        0.276662 -0.472035
3 cat    long        -0.013960 -0.362543
  dog    short       -0.006154 -0.923061

However if the subgroups do not have the same set of labels i.e long/short or cat/dog then there might be missing data when using stack and unstack methods. i.e.

exp                  B          
animal             dog       cat
first second                    
bar   one     0.805244 -1.206412
      two     1.340309 -1.170299
foo   one     1.607920  1.024180
qux   two     0.769804 -1.281247

When you unstack this using df.unstack()

exp            B                              
animal       dog                 cat          
second       one       two       one       two
first                                         
bar     0.805244  1.340309 -1.206412 -1.170299
foo     1.607920       NaN  1.024180       NaN
qux          NaN  0.769804       NaN -1.281247

To overcome this issue, you can use the fill_value parameter to modify what the missing values will be set as. It is important that missing values are handled as soon as they show up to avoid it affecting later usage of the data frame causing unexpected behavior.