pandas describe by - additional parameters

Question

I see that the pandas library has a Describe by function which returns some useful statistics. However, is there a way to add additional rows to the output such as standard deviation (.std) and median absolute deviation (.mad) or the count of unique values?

I get df.describe() but I'm unable to find out how to add these additional summary things

Answer 1

the default describe looks like this:

np.random.seed([3,1415])
df = pd.DataFrame(np.random.rand(100, 5), columns=list('ABCDE'))

df.describe()

                A           B           C           D           E
count  100.000000  100.000000  100.000000  100.000000  100.000000
mean     0.495871    0.472939    0.455570    0.503899    0.451341
std      0.303589    0.291968    0.294984    0.269936    0.284666
min      0.006453    0.001559    0.001068    0.015311    0.009526
25%      0.239379    0.219141    0.196251    0.294371    0.202956
50%      0.529596    0.456548    0.376558    0.532002    0.432936
75%      0.759452    0.739666    0.665563    0.730702    0.686793
max      0.999799    0.994510    0.997271    0.981551    0.979221

Updated for pandas > 0.21.0
I'd make my own describe like below. It should be obvious how to add more.

def describe(df, stats):
    d = df.describe()
    return d.append(df.reindex(d.columns, axis = 1).agg(stats))

describe(df, ['skew', 'mad', 'kurt'])

                A           B           C           D           E
count  100.000000  100.000000  100.000000  100.000000  100.000000
mean     0.495871    0.472939    0.455570    0.503899    0.451341
std      0.303589    0.291968    0.294984    0.269936    0.284666
min      0.006453    0.001559    0.001068    0.015311    0.009526
25%      0.239379    0.219141    0.196251    0.294371    0.202956
50%      0.529596    0.456548    0.376558    0.532002    0.432936
75%      0.759452    0.739666    0.665563    0.730702    0.686793
max      0.999799    0.994510    0.997271    0.981551    0.979221
skew    -0.014942    0.048054    0.247244   -0.125151    0.066156
mad      0.267730    0.249968    0.254351    0.228558    0.242874
kurt    -1.323469   -1.223123   -1.095713   -1.083420   -1.148642

Updated for pandas 0.20
I'd make my own describe like below. It should be obvious how to add more.

def describe(df, stats):
    d = df.describe()
    return d.append(df.reindex_axis(d.columns, 1).agg(stats))

describe(df, ['skew', 'mad', 'kurt'])

                A           B           C           D           E
count  100.000000  100.000000  100.000000  100.000000  100.000000
mean     0.495871    0.472939    0.455570    0.503899    0.451341
std      0.303589    0.291968    0.294984    0.269936    0.284666
min      0.006453    0.001559    0.001068    0.015311    0.009526
25%      0.239379    0.219141    0.196251    0.294371    0.202956
50%      0.529596    0.456548    0.376558    0.532002    0.432936
75%      0.759452    0.739666    0.665563    0.730702    0.686793
max      0.999799    0.994510    0.997271    0.981551    0.979221
skew    -0.014942    0.048054    0.247244   -0.125151    0.066156
mad      0.267730    0.249968    0.254351    0.228558    0.242874
kurt    -1.323469   -1.223123   -1.095713   -1.083420   -1.148642

Old Answer

def describe(df):
    return pd.concat([df.describe().T,
                      df.mad().rename('mad'),
                      df.skew().rename('skew'),
                      df.kurt().rename('kurt'),
                     ], axis=1).T

describe(df)

                A           B           C           D           E
count  100.000000  100.000000  100.000000  100.000000  100.000000
mean     0.495871    0.472939    0.455570    0.503899    0.451341
std      0.303589    0.291968    0.294984    0.269936    0.284666
min      0.006453    0.001559    0.001068    0.015311    0.009526
25%      0.239379    0.219141    0.196251    0.294371    0.202956
50%      0.529596    0.456548    0.376558    0.532002    0.432936
75%      0.759452    0.739666    0.665563    0.730702    0.686793
max      0.999799    0.994510    0.997271    0.981551    0.979221
mad      0.267730    0.249968    0.254351    0.228558    0.242874
skew    -0.014942    0.048054    0.247244   -0.125151    0.066156
kurt    -1.323469   -1.223123   -1.095713   -1.083420   -1.148642

Answer 2

The answer from piRSquared makes the most sense to me, but I get a deprecation warning about reindex_axis in Python 3.5. This works for me:

    stats = data.describe()
    stats.loc['IQR'] = stats.loc['75%'] - stats.loc['25%'] # appending interquartile range instead of recalculating it
    stats = stats.append(data.reindex(stats.columns, axis=1).agg(['skew', 'mad', 'kurt']))