熊猫数据框的条件选择

Question

我有给定格式的数据集

   time    color height weight value
1  t1      red    hr1     wr1     vr1
2  t1      red    hr1     wr1     vr1
3  t1      blue   hb1     wb1    vb1
4  t1      blue   hb1     wb1     vb1
5  t1      green  hg1     wg1     vg1
6  t1      green  hg1     wg1     vg1
7  t2      blue   hb2     wb2     vb2
8  t2      green  hg2     wg2    vg2
9  t2      red    hr2     wr2     vr2
10 t2      red    hr2     wr2     vr2
11 t3      red    hr3     wr3     vr3
12 t3      red    hr3     wr3     vr3
13 t3      green  hg3     wg3     vg3
14 t3      green  hg3     wg3     vg3
15 t3      blue   hb3     wb3     vb3
16 t3      blue   hb3     wb3     vb3

我想放弃时间的测量，因为对于每种红色，蓝色和绿色，颜色的计数值都不相同。 在给定的代码段中，应保留t1和t3，并删除所有用于t3测量的行。

结果应为：

 time   color height weight value
1  t1      red    hr1     wr1     vr1
2  t1      red    hr1     wr1     vr1
3  t1      blue   hb1     wb1    vb1
4  t1      blue   hb1     wb1     vb1
5  t1      green  hg1     wg1     vg1
6  t1      green  hg1     wg1     vg1
7  t3      red    hr3     wr3     vr3
8  t3      red    hr3     wr3     vr3
9  t3      green  hg3     wg3     vg3
10 t3      green  hg3     wg3     vg3
11  t3     blue   hb3     wb3     vb3
12  t3     blue   hb3     wb3     vb3

谢谢

Answer 1

怎么样：

s = df.groupby(['time', 'color']).size()
s = s.unstack(0).eq(2).all()
valid_times = s.index[s]

print(df[df.time.isin(valid_times)])

   time  color height weight value
1    t1    red    hr1    wr1   vr1
2    t1    red    hr1    wr1   vr1
3    t1   blue    hb1    wb1   vb1
4    t1   blue    hb1    wb1   vb1
5    t1  green    hg1    wg1   vg1
6    t1  green    hg1    wg1   vg1
11   t3    red    hr3    wr3   vr3
12   t3    red    hr3    wr3   vr3
13   t3  green    hg3    wg3   vg3
14   t3  green    hg3    wg3   vg3
15   t3   blue    hb3    wb3   vb3
16   t3   blue    hb3    wb3   vb3

Answer 2

对返回系列使用双精度GroupBy.transform，其大小与原始DataFrame相同，因此可以使用boolean indexing：

df1 = df[df.groupby(['time', 'color'])['color']
           .transform('size')
           .eq(2)
           .groupby(df['time'])
           .transform('all')]

print (df1)
   time  color height weight value
1    t1    red    hr1    wr1   vr1
2    t1    red    hr1    wr1   vr1
3    t1   blue    hb1    wb1   vb1
4    t1   blue    hb1    wb1   vb1
5    t1  green    hg1    wg1   vg1
6    t1  green    hg1    wg1   vg1
11   t3    red    hr3    wr3   vr3
12   t3    red    hr3    wr3   vr3
13   t3  green    hg3    wg3   vg3
14   t3  green    hg3    wg3   vg3
15   t3   blue    hb3    wb3   vb3
16   t3   blue    hb3    wb3   vb3