How to select values that have 1:1 indexes in a datataframe with multiindex?

Question

I have a dataframe with a multi-index an I want to select only those rows whose indexes are 1:1

arrays = [np.array(['bar', 'bar', 'baz', 'baz', 'foo', 'foo', 'qux']),
          np.array(['one', 'two', 'one', 'two', 'one', 'two', 'one' ])]

df = pd.DataFrame(np.random.randn(7, 4), index=arrays)

I expect to select only

qux one -0.521636 -2.253847 -0.420486 1.354772

Answer 1

I believe you need remove duplicated rows by first level of MultiIndex with Index.get_level_values, Index.duplicated and boolean indexing:

df = df[~df.index.get_level_values(0).duplicated(keep=False)]
print (df)
                0         1         2         3
qux one -1.031044 -0.838885  0.316934  0.353254

Or use GroupBy.transform with any column - e.g. first (or you can specify it) for counts per first level:

df = df[df.groupby(level=0)[df.columns[0]].transform('size').eq(1)]