CODEWITHSUNDEEP
pythonpandas
Laura
Laura

Reputation: 191

Python, pandas: How to append a series to a dataframe

I have the following dataframe df1: 

import pandas as pd
data = {'name': ['Jason', 'Molly', 'Tina', 'Jake', 'Amy', 'Lisa', 'Molly', 'Lisa', 'Molly', 'Fred'], 
             'gender': ['m', 'f', 'f', 'm', 'f', 'f', 'f', 'f','f', 'm'], 
   }
df1 = pd.DataFrame(data, index = [1, 2, 3, 4, 5, 6, 7, 8, 9, 10])

I want to create a table with some standard and some custom summary statistics df2. 

df2 = df1.describe()
df2.rename(index={'top':'mode'},inplace=True)
df2.rename(index={'freq':'mode freq'},inplace=True)
df2

df2:

                  gender    name
        count       10      10
        unique      2       7
        mode        f       Molly
        mode freq   7       3

I want to append one row to df2 for the second mode and one for the frequency of the second mode:

Example: 

                gender  name
    count       10      10
    unique      2       7
    mode        f       Molly
    mode freq   7       3
    2nd mode    m       Lisa
    2nd freq    3       2

I figured out that you can get the second mode & frequency by doing this: 

my_series
for column in df1:
   my_series=df1[column].value_counts()[1:2]
   print(my_series)

But how do I append this to df2?

Upvotes: 3

Views: 296

Answers (2)

piRSquared
piRSquared

Reputation: 294258

With Counter

from collections import Counter

def f(s):
    return pd.Series(Counter(s).most_common(2)[1], ['mode2', 'mode2 freq'])

df1.describe().rename(dict(top='mode1', freq='mode1 freq')).append(df1.apply(f))

             name gender
count          10     10
unique          7      2
mode1       Molly      f
mode1 freq      3      7
mode2        Lisa      m
mode2 freq      2      3

value_counts

Same thing without Counter

def f(s):
    c = s.value_counts()
    return pd.Series([s.iat[1], s.index[1]], ['mode2', 'mode2 freq'])

df1.describe().rename(dict(top='mode1', freq='mode1 freq')).append(df1.apply(f))

Numpy bits

def f(s):
    f, u = pd.factorize(s)
    c = np.bincount(f)
    i = np.argpartition(c, -2)[-2]
    return pd.Series([u[i], c[i]], ['mode2', 'mode2 freq'])

df1.describe().rename(dict(top='mode1', freq='mode1 freq')).append(df1.apply(f))

Upvotes: 2

BENY
BENY

Reputation: 323226

You can do apply with value_counts, then we need modify your dataframe shape . 

df1.apply(lambda x : x.value_counts().iloc[[1]]).stack().reset_index(level=0).T
Out[172]: 
         name gender
level_0  Lisa      m
0           2      3

The final out put (Change the index name using what you show to us rename)

pd.concat([df1.describe(),df1.apply(lambda x : x.value_counts().iloc[[1]]).stack().reset_index(level=0).T])
Out[173]: 
        gender   name
count       10     10
unique       2      7
top          f  Molly
freq         7      3
level_0      m   Lisa
0            3      2

Upvotes: 4

Related Questions