I have collected the frequencies in which combinations of certain categorical parameters occur in a DataFrame. Applying a groupby operation produced another DataFrame with a MultiIndex. Now I would like to visualize the frequencies in a mosaic plot. This is what I have tried:

rows=[
    {"Mode":"ID", "SortBy":"Start", "SortDir":"ASC", "count":10},
    {"Mode":"ID", "SortBy":"Start", "SortDir":"DESC", "count":100},
    {"Mode":"FULL", "SortBy":"End", "SortDir":"DESC", "count":1000}
]
df=pd.DataFrame(rows)
hdf=df.groupby(["Mode", "SortBy", "SortDir"]).sum("count")
mosaic(hdf, index=hdf.index)

But it fails with the following error:

KeyError: "None of [MultiIndex([('FULL',   'End', 'DESC'),n            (  'ID', 'Start',  'ASC'),n            (  'ID', 'Start', 'DESC')],n           names=['Mode', 'SortBy', 'SortDir'])] are in the [columns]"

I was able to produce a diagram using

mosaic(hdf, ["count"])

But this is obviously not what I want: it shows three same-sized rectangles labelled 10, 100, 1000, whereas I was expecting the categories Mode, SortBy, SortDir arranged around the axes and the rectangles reflecting the proportions of counts.