如果我有这两栏:
dat=[['yes','dog', 20,4,60,400],['yes','dog', 20,4,60,300],['yes','cat', 20,10,10,float('nan')]]
df_dat= pd.DataFrame(dat,columns = ["Time","animal", "val", "val2", "val3", "val4"])
我想获得一个使用“时间”和“动物”分组的数据框。然后,它采用其他列的组合方式。一个子集是[“ val”,“ val3”]和[“ val2”,“ val4”]。
基本上,某些东西可以利用df_dat.groupby([“ Time”,“ animal”])。mean()的值列子集的结果
我正在寻找的输出看起来像(但采用数据帧格式):
[Index , 'val'/'val3','val2/val4']
[('yes','dog'),40,177]
[('yes','cat'),15,10]
答案 0 :(得分:1)
我相信您需要
ndf = df_dat.groupby(['Time', 'animal']).mean()
ndf['v1v3'], ndf['v2v4'] = ndf[['val', 'val3']].mean(1), ndf[['val2', 'val4']].mean(1)
输出
val val2 val3 val4 v1v3 v2v4
Time animal
yes cat 20 10 10 NaN 15.0 10.0
dog 20 4 60 350.0 40.0 177.0
当然可以选择均值列
ndf[['v1v3', 'v2v4']]
v1v3 v2v4
Time animal
yes cat 15.0 10.0
dog 40.0 177.0
答案 1 :(得分:1)
设置
df = df_dat.groupby(['Time', 'animal']).mean()
subsets = [["val","val3"], ["val2","val4"]]
使用字典理解和 assign
:
df.assign(**{'/'.join(cols): df[cols].mean(1) for cols in subsets})
val val2 val3 val4 val/val3 val2/val4
Time animal
yes cat 20 10 10 NaN 15.0 10.0
dog 20 4 60 350.0 40.0 177.0
如果您只想要子集列:
pd.DataFrame({'/'.join(cols): df[cols].mean(1) for cols in subsets})
val/val3 val2/val4
Time animal
yes cat 15.0 10.0
dog 40.0 177.0