Question

I’ve managed to make the following bar plot using matplotlib.pyplot.

The plot comes from an aggregated PANDAS DataFrame, printed below. Note that each bar in the bar-plot corresponds to a value in the mean column. Also note that the values are not zero, but that PANDAS outputs 0 and -0 when floats are very small.

Group Local Global Attn  mean
ASub  LD    GD     Attn    -0
                   Dist    -0
            GS     Attn    -0
                   Dist    -0
      LS    GD     Attn    -0
                   Dist    -0
            GS     Attn    -0
                   Dist    -0
DSub  LD    GD     Attn    -0
                   Dist     0
            GS     Attn    -0
                   Dist    -0
      LS    GD     Attn    -0
                   Dist    -0
            GS     Attn    -0
                   Dist    -0

I would like to label the x-axis of my bar plot hierarchically, in a manner corresponding to the labels above. In other words, the left half of the x-axis corresponds to the ASub group. The left half of the ASub group corresponds to the LD level of the Local factor, and so on…

Can this be done?

EDIT:

I think I should probably clarify exactly what I want. I’d like for there to be several labels, progressing from most general (Group) to most specific (Attn), similar to the organization on the left 4 columns of the DataFrame above.

Answer 1

On master this is implemented (i used random data for ‘mean’), see image at the bottom.
If you prefer not to upgrade you can also set it manually:

In [143]: ax = df.plot(kind='bar')

In [144]: ax.set_xticklabels(['|'.join(t) for t in df.index])

EDIT:

 In [167]: ax = df.plot(kind='bar')

 In [168]: ax.set_xticklabels(df.index.format(names=False))