Question

我想在groupBy对象上计算多个值（包含在每个单元格的列表中）。

我有以下数据框：

| | Record the respondent’s sex | 7. What do you use the phone for? | |---|-----------------------------|---------------------------------------------| | 0 | Male | sending texts;calls;receiving sending texts | | 1 | Female | sending texts;calls;WhatsApp;Facebook | | 2 | Male | sending texts;calls;receiving texts | | 3 | Female | sending texts;calls |

在7. What do you use the phone for?分组后，我想计算Record the respondent’s sex列中的每个值。

当每个单元格只有一个值时，我没有问题。

grouped = df.groupby(['Record the respondent’s sex'], sort=True)

question_counts = grouped['2. Are you a teacher, caregiver, or young adult ?'].value_counts(normalize=False, sort=True)

question_data = [
      {'2. Are you a teacher, caregiver, or young adult ?': question, 'Record the respondent’s sex': group, 'count': count*100} for
      (group, question), count in dict(question_counts).items()]

df_question = pd.DataFrame(question_data)

给我一张看起来完全像这样的表格：

| 7. What do you use the phone for? | Record the respondent's sex | count | |-----------------------------------|-----------------------------|-------| | sending texts | Male | 2 | | calls | Male | 2 | | receiving texts | Male | 2 | | sending texts | Female | 2 | | calls | Female | 2 | | WhatsApp | Female | 1 | | Facebook | Female | 1 |

如果只有我可以使用多个值！

value_counts()无法处理具有多个值的列表，会引发TypeError: unhashable type: 'list'错误。问题Counting occurrence of values in a Panda series?显示了如何以各种方式处理这个问题，但我似乎无法让它在GroupBy对象上工作。

Answer 1

# Initialize sample data.
df = pd.DataFrame({'Record the respondent’s sex': ['Male', 'Female'] * 2, 
                   '7. What do you use the phone for?': [
                       "sending texts;calls;receiving sending texts",
                       "sending texts;calls;WhatsApp;Facebook",
                       "sending texts;calls;receiving texts",
                       "sending texts;calls"
                   ]})

# Split the values on ';' and separate into columns.  Melt the result.
df2 = pd.melt(
    pd.concat([df['Record the respondent’s sex'],
               df.loc[:, "7. What do you use the phone for?"].apply(
                   lambda series: series.split(';')).apply(pd.Series)], axis=1),
    id_vars='Record the respondent’s sex')[['Record the respondent’s sex', 'value']]

# Group on gender and rename columns.
result = df2.groupby('Record the respondent’s sex')['value'].value_counts().reset_index()
result.columns = ['Record the respondent’s sex', '7. What do you use the phone for?', 'count']

# Reorder columns.
>>> result[['7. What do you use the phone for?', 'Record the respondent’s sex', 'count']]
  7. What do you use the phone for? Record the respondent’s sex  count
0                             calls                      Female      2
1                     sending texts                      Female      2
2                          Facebook                      Female      1
3                          WhatsApp                      Female      1
4                             calls                        Male      2
5                     sending texts                        Male      2
6           receiving sending texts                        Male      1
7                   receiving texts                        Male      1

计算groupby对象中的多个值

1 个答案: