Python 带熊猫的过滤组

我正在尝试为使用Pandas的组添加过滤器。在下面的棒球数据中，我希望计算出在“入选”列中从初始“N”到最终“Y”所需的平均时间。基本上，我希望计算每个组的长度，这些组在导入列中包含一个“Y”，并且有多行。任何提示都会有帮助

   playerID  yearid votedBy  ballots  needed  votes inducted category needed_note
2860  aaronha01    1982   BBWAA      415     312    406        Y   Player         NaN
3743  abbotji01    2005   BBWAA      516     387     13        N   Player         NaN
 146  adamsba01    1937   BBWAA      201     151      8        N   Player         NaN
 259  adamsba01    1938   BBWAA      262     197     11        N   Player         NaN
 384  adamsba01    1939   BBWAA      274     206     11        N   Player         NaN
 497  adamsba01    1942   BBWAA      233     175     11        N   Player         NaN
 574  adamsba01    1945   BBWAA      247     186      7        N   Player         NaN
2108  adamsbo03    1966   BBWAA      302     227      1        N   Player         NaN

---

我修改了您的数据集，以便有两个这样的组。一个从

N

到

Y

有2行，另一个从

N

到

Y

有8行。这取决于是否在包含行的

y

中计数。如果没有，它将有两个组，一个包含1行，另一个包含7行。它看起来也没有时间序列列，所以我猜这意味着行在时间上是均匀分布的

In [25]:

df=pd.read_clipboard()
print df
       playerID  yearid votedBy  ballots  needed  votes inducted category  needed_note 
3741  abbotji01    2005   BBWAA      516     387     13        N   Player          NaN 
2860  aaronha01    1982   BBWAA      415     312    406        Y   Player          NaN 
3743  abbotji01    2005   BBWAA      516     387     13        N   Player          NaN 
146   adamsba01    1937   BBWAA      201     151      8        N   Player          NaN 
259   adamsba01    1938   BBWAA      262     197     11        N   Player          NaN 
384   adamsba01    1939   BBWAA      274     206     11        N   Player          NaN 
497   adamsba01    1942   BBWAA      233     175     11        N   Player          NaN 
574   adamsba01    1945   BBWAA      247     186      7        N   Player          NaN 
2108  adamsbo03    1966   BBWAA      302     227      1        N   Player          NaN 
2861  aaronha01    1982   BBWAA      415     312    406        Y   Player          NaN 

In [26]:

df['isY']=(df.inducted=='Y')
df['isY']=np.hstack((0,df['isY'].cumsum().values[:-1])).T
In [27]:

print df.groupby('isY').count()
     playerID  yearid  votedBy  ballots  needed  votes  inducted  category  needed_note  isY 
0           2       2        2        2       2      2         2         2            0    2 
1           8       8        8        8       8      8         8         8            0    8 
[2 rows x 10 columns]    

假设您不计算

Y

，则可通过以下方式计算平均值：

df2=df.groupby('isY').count().isY-1
df2[df2!=1].mean()

---

我模拟了我自己的数据，为你的问题制作了一个简单的测试用例。我创建了一组名为DFIL的玩家，包括最终被诱导的玩家，然后使用ISIN（）函数，我们可以确保只在分析中考虑它们。然后我找到他们约会的最小值和最大值，并平均它们的差异

import pandas as pd

df = pd.DataFrame({'player':['Nate','Will','Nate','Will'], 
                   'inducted': ['Y','Y','N','N'],
                   'date':[2014,2000,2011,1999]})

df_inducted = df[df.inducted=='Y']
df_subset = df[df.player.isin(df_inducted.player)]

maxs = df_subset.groupby('player')['date'].max()
mins = df_subset.groupby('player')['date'].min()

maxs = pd.DataFrame(maxs)
maxs.columns = ['max_date']
mins = pd.DataFrame(mins)
mins.columns = ['min_date']

min_and_max = maxs.join(mins)
final = min_and_max['max_date'] - min_and_max['min_date']

print "average time:", final.mean()

---

类DataFrameGroupBy的filter方法对组中的每个子帧进行操作。请参阅

帮助（pd.core.groupby.DataFrameGroupBy.filter）

。数据如下：

print df
  inducted playerID
0        Y        a
1        N        a
2        N        a
3        Y        b
4        N        b
5        N        c
6        N        c
7        N        c

示例代码：

import pandas as pd

g = df.groupby('playerID')
madeit = g.filter(
        lambda subframe:
                'Y' in set(subframe.inducted)).groupby('playerID')

# The filter removed player 'c' who never has inducted == 'Y'
print madeit.head()
           inducted playerID
playerID                    
a        0        Y        a
         1        N        a
         2        N        a
b        3        Y        b
         4        N        b

# The 'aggregate' function applies a function to each subframe
print madeit.aggregate(len)
          inducted
playerID          
a                3
b                2