Pandas: change between mean/std and plus/minus notations

Question

Let's assume I have a Pandas's DataFrame:

import numpy as np
import pandas as pd

df = pd.DataFrame(
    np.random.randint(0, 100, size=(10, 4)), columns=('A', 'DA', 'B', 'DB'))

which outputs:

    A  DA   B  DB
0  62  87  10  39
1  56   7  81  12
2  37  26  21  44
3  56  26  42  32
4  29  45  11   9
5  11  85   4  79
6  87  31  61  90
7   5  55  26  47
8  55  94  20  84
9  52  26  72  19

I would like to convert that to this:

       A      B
0  62±87  10±39
1   56±7  81±12
2  37±26  21±44
3  56±26  42±32
4  29±45   11±9
5  11±85   4±79
6  87±31  61±90
7   5±55  26±47
8  55±94  20±84
9  52±26  72±19

and viceversa.

I could do this "by hand" but I was hoping for an elegant way using Pandas' built-ins, which could eventually be converted elegantly to LaTeX (i.e. 62±87 becomes $62 \pm 87$).

I was looking into .apply() from Converting a column within pandas dataframe from int to string but it is not clear to me how to use it for this purpose.

EDIT

The suggested answer do NOT seem to cover the VICEVERSA: i.e. convertin from the A±DA notation back to two columns A and DA.

Answer 1

Here's one way

In [1336]: (df.groupby(df.columns.str[-1], axis=1)
              .apply(lambda x: x.astype(str).apply('±'.join, 1)))
Out[1336]:
       A      B
0  62±87  10±39
1   56±7  81±12
2  37±26  21±44
3  56±26  42±32
4  29±45   11±9
5  11±85   4±79
6  87±31  61±90
7   5±55  26±47
8  55±94  20±84
9  52±26  72±19

Another way

In [1351]: pd.DataFrame({c: df.filter(like=c).astype(str).apply('±'.join, 1) 
                         for c in df.columns.str[-1].unique()})
Out[1351]:
       A      B
0  62±87  10±39
1   56±7  81±12
2  37±26  21±44
3  56±26  42±32
4  29±45   11±9
5  11±85   4±79
6  87±31  61±90
7   5±55  26±47
8  55±94  20±84
9  52±26  72±19

Or, also as

In [1386]: pd.DataFrame({c: ['±'.join(v) for v in df.filter(like='A').astype(str).values]
      ...:               for c in df.columns.str[-1].unique()})

---

And, opposite assuming dff is your string joined dataframe

In [1357]: pd.concat([dff[c].str.split('±', expand=True).rename(columns={0:c, 1:'D'+c})
                      for c in dff.columns], axis=1)
Out[1357]:
    A  DA   B  DB
0  62  87  10  39
1  56   7  81  12
2  37  26  21  44
3  56  26  42  32
4  29  45  11   9
5  11  85   4  79
6  87  31  61  90
7   5  55  26  47
8  55  94  20  84
9  52  26  72  19

---

Details

In [1358]: df
Out[1358]:
    A  DA   B  DB
0  62  87  10  39
1  56   7  81  12
2  37  26  21  44
3  56  26  42  32
4  29  45  11   9
5  11  85   4  79
6  87  31  61  90
7   5  55  26  47
8  55  94  20  84
9  52  26  72  19

In [1359]: dff
Out[1359]:
       A      B
0  62±87  10±39
1   56±7  81±12
2  37±26  21±44
3  56±26  42±32
4  29±45   11±9
5  11±85   4±79
6  87±31  61±90
7   5±55  26±47
8  55±94  20±84
9  52±26  72±19

Helpers

In [1377]: df.columns.str[-1]
Out[1377]: Index([u'A', u'A', u'B', u'B'], dtype='object')

In [1378]: df.columns.str[-1].unique()
Out[1378]: Index([u'A', u'B'], dtype='object')

Answer 2

Here I found a bunch of them so possible a duplicate: Combine two columns of text in dataframe in pandas/python

This one convinced me the most:

import io
import pandas as pd

string = """A,DA,B,DB
62,87,10,39"""

df = pd.read_csv(io.StringIO(string),sep=",")

cols = [i for i in df.columns if len(i) == 1]

for i in cols:
    df[i] = df[i].astype(str)+ "±" + df["D"+i].astype(str)

df[cols]