大熊貓：計算從所述差的分組的平均

Question

我具有由月和年爲多個傳感器的傳感器數據：

import pandas as pd 
df = pd.DataFrame([ 
['A', 'Jan', 2015, 13], 
['A', 'Feb', 2015, 10], 
['A', 'Jan', 2016, 12], 
['A', 'Feb', 2016, 11], 
['B', 'Jan', 2015, 7], 
['B', 'Feb', 2015, 8], 
['B', 'Jan', 2016, 4], 
['B', 'Feb', 2016, 9] 
], columns = ['sensor', 'month', 'year', 'value']) 

In [2]: df 
Out[2]: 
    sensor month year value 
0  A Jan 2015  13 
1  A Feb 2015  10 
2  A Jan 2016  12 
3  A Feb 2016  11 
4  B Jan 2015  7 
5  B Feb 2015  8 
6  B Jan 2016  4 
7  B Feb 2016  9 

予算出的平均每個傳感器和每月用GROUPBY：

month_avg = df.groupby(['sensor', 'month']).mean()['value'] 

In [3]: month_avg 
Out[3]: 
sensor month 
A  Feb  10.5 
     Jan  12.5 
B  Feb  8.5 
     Jan  5.5 

現在我想添加一列到df與月平均值的差異，如下所示：

sensor month year value diff_from_avg 
0  A Jan 2015  13 1.5 
1  A Feb 2015  10 2.5 
2  A Jan 2016  12 0.5 
3  A Feb 2016  11 0.5 
4  B Jan 2015  7 2.5 
5  B Feb 2015  8 0.5 
6  B Jan 2016  4 -1.5 
7  B Feb 2016  9 -0.5 

我想多索引df和avgs_by_month相似，並試圖簡單的減法，但沒有好：

df = df.set_index(['sensor', 'month']) 
df['diff_from_avg'] = month_avg - df.value 

謝謝你的任何建議。

Answer 1

assign與transform

diff_from_avg=df.value - df.groupby(['sensor', 'month']).value.transform('mean') 
df.assign(diff_from_avg=diff_from_avg) 

    sensor month year value diff_from_avg 
0  A Jan 2015  13   0.5 
1  A Feb 2015  10   -0.5 
2  A Jan 2016  12   -0.5 
3  A Feb 2016  11   0.5 
4  B Jan 2015  7   1.5 
5  B Feb 2015  8   -0.5 
6  B Jan 2016  4   -1.5 
7  B Feb 2016  9   0.5 

Answer 2

您需要設置數據框的索引與分組系列一致的新列，那麼你可以直接減去：

df.set_index(['sensor','month'], inplace=True) df['diff'] = df['value'] - month_avg

Answer 3

嘗試：

df['diff_from_avg']=df.groupby(['sensor','month'])['value'].apply(lambda x: x-x.mean()) 
Out[18]: 
    sensor month year value diff_from_avg 
0  A Jan 2015  13   0.5 
1  A Feb 2015  10   -0.5 
2  A Jan 2016  12   -0.5 
3  A Feb 2016  11   0.5 
4  B Jan 2015  7   1.5 
5  B Feb 2015  8   -0.5 
6  B Jan 2016  4   -1.5 
7  B Feb 2016  9   0.5