在 Python 中使用 multiindex grouped_by 数据框对象生成随机数据

Question

下表包含有关每个领导者和费用类型的费用的汇总统计信息。我将稳定存储在 python 中作为多索引数据框对象。我的目标是使用每个类别下的平均值和标准差为每个领导者和费用类型生成随机数据（运行下面的代码 sn-p 以获取表格）。有一个“计数”列，表示我想为每个 Leader-Expense_Type 组合生成多少个随机数。我提出了广泛而低效的循环结构，似乎无法正确完成工作。我应该如何解决这个问题？

注意：这只是数据的一个样本。还有更多的领导者拥有同样多的费用类型。

<table border="1" class="dataframe">  <thead>    <tr>      <th></th>      <th></th>      <th colspan="3" halign="left">Expense_Amount</th>    </tr>    <tr>      <th></th>      <th></th>      <th>mean</th>      <th>std</th>      <th>count</th>    </tr>    <tr>      <th>Leader</th>      <th>Expense_Type</th>      <th></th>      <th></th>      <th></th>    </tr>  </thead>  <tbody>    <tr>      <th rowspan="7" valign="top">Leader1</th>      <th>Airfare</th>      <td>1979.684219</td>      <td>2731.629767</td>      <td>1358</td>    </tr>    <tr>      <th>Booking Fees</th>      <td>118.994538</td>      <td>270.007390</td>      <td>1179</td>    </tr>    <tr>      <th>Conference/Seminars</th>      <td>1553.830923</td>      <td>1319.295946</td>      <td>65</td>    </tr>    <tr>      <th>Hotel</th>      <td>1656.643658</td>      <td>2104.721093</td>      <td>1405</td>    </tr>    <tr>      <th>Meals</th>      <td>435.665122</td>      <td>676.705857</td>      <td>1476</td>    </tr>    <tr>      <th>Mileage</th>      <td>213.785046</td>      <td>284.908031</td>      <td>979</td>    </tr>    <tr>      <th>Taxi/Uber</th>      <td>308.530724</td>      <td>380.288964</td>      <td>1422</td>    </tr>    <tr>      <th rowspan="7" valign="top">Leader2</th>      <th>Airfare</th>      <td>1730.196911</td>      <td>2334.688155</td>      <td>628</td>    </tr>    <tr>      <th>Booking Fees</th>      <td>112.020556</td>      <td>573.407269</td>      <td>576</td>    </tr>    <tr>      <th>Conference/Seminars</th>      <td>1647.576500</td>      <td>1154.320584</td>      <td>80</td>    </tr>    <tr>      <th>Hotel</th>      <td>1693.080356</td>      <td>1953.552474</td>      <td>618</td>    </tr>    <tr>      <th>Meals</th>      <td>574.228548</td>      <td>844.997595</td>      <td>620</td>    </tr>    <tr>      <th>Mileage</th>      <td>215.898798</td>      <td>291.231331</td>      <td>466</td>    </tr>    <tr>      <th>Taxi/Uber</th>      <td>298.655852</td>      <td>340.926518</td>      <td>569</td>    </tr>  </tbody></table>

Answer 1

您可以将df.apply(your_function, axis=1) 用于

def your_function(df):
    mean = df['mean']
    std = df['std']
    result = mean  # Replace with your number generator
    return result

有关更广泛的描述，请参阅此答案：How to apply a function to two columns of Pandas dataframe

Answer 2

这是我的解决方案：

# Dictionary to hold generated data
rand_expenses_dict = {}

# Loop over each unique leader
for leader in agg_data.index.get_level_values("Leader").unique():

# Loop over each unique expense type
for expense_type in agg_data.index.get_level_values("Expense_Type").unique():

    # Not al leaders have all expense types
    # The exception handling method will ignore expense types
    # That do not correspond to a leader
    try:

        # Generate random numbers
        rand = (np.round(
                        np.random.normal(
                            loc=agg_data.loc[leader, expense_type][0],
                            scale = agg_data.loc[leader, expense_type][1],
                            size  = int(agg_data.loc[leader, expense_type][2])
                        ),2))

        # Add random numbers to data dictionaty
        rand_expenses_dict[(leader,expense_type)] = rand

    # If it finds an error, go to the next expense
    except:
        pass