pandas 具有多索引的高级横截面 [英] Advanced cross-section with multi-index in pandas
本文介绍了 pandas 具有多索引的高级横截面的处理方法,对大家解决问题具有一定的参考价值,需要的朋友们下面随着小编来一起学习吧!
问题描述
我有以下数据框:
lb = [('A','a',1), ('A','a',2), ('A','a',3), ('A','b',1), ('A','b',2), ('A','b',3), ('B','a',1), ('B','a',2), ('B','a',3), ('B', 'b',1), ('B','b',2) ,('B','b',3)]
col = pd.MultiIndex.from_tuples(lb, names=['first','second','third'])
df = pd.DataFrame(randn(5,12), columns=col)
first A B \
second a b a
third 1 2 3 1 2 3 1
0 1.597958 2.054695 0.449745 -0.990393 0.780978 -0.590558 -0.691706
1 -0.093841 -1.203769 1.779555 -0.299931 -0.411360 0.122852 -0.250156
2 0.025183 0.514480 -0.420666 1.574669 0.962010 1.278237 -0.976286
3 -1.028288 -0.506581 0.880370 1.513487 -0.066479 -0.100231 0.785042
4 -1.635642 0.464074 -0.335941 -0.034194 0.412519 -0.672058 0.113886
first
second b
third 2 3 1 2 3
0 1.954769 0.705860 -1.712058 1.015807 1.245232
1 -2.037299 -0.120649 -0.114652 -0.686707 -0.993540
2 0.918084 -0.892378 -0.741131 -2.547121 0.797637
3 0.000077 2.123063 0.903571 1.972190 -1.179325
4 -1.145241 -1.773182 0.407046 -0.301640 -0.173261
我想获取2和3的所有列,即
I want to obtain all columns with 2 and 3, that is, sth like
df.xs([2,3], level='third', axis=1, drop_level=False)
但这不起作用.我该如何进行?
But this doesn't work. How do I proceed?
推荐答案
This is a new feature in 0.14.0, see whatsnew here. This effectively replaces the need for .xs
.
In [8]: idx = pd.IndexSlice
In [9]: df.loc[:,idx[:,:,[2,3]]]
Out[9]:
first A B
second a b a b
third 2 3 2 3 2 3 2 3
0 1.770120 -0.362269 -0.804352 1.549652 0.069858 -0.274113 0.570410 -0.460956
1 -0.982169 2.044497 0.571353 0.310634 -1.865966 -0.862613 0.124413 0.645419
2 -1.412519 0.168448 0.081467 -0.220464 1.033748 1.561429 0.094363 0.254768
3 -0.653458 -0.978661 0.158708 -0.818675 -1.122577 0.026941 2.678548 0.864817
4 -0.555179 -0.155564 1.148956 1.438523 -1.254660 0.609254 -0.970612 1.519028
要减去,这是不平凡的.
To subtract this is non-trivial.
[107]: df = pd.DataFrame(np.arange(5*12).reshape(-1,12), columns=col)
In [108]: df
Out[108]:
first A B
second a b a b
third 1 2 3 1 2 3 1 2 3 1 2 3
0 0 1 2 3 4 5 6 7 8 9 10 11
1 12 13 14 15 16 17 18 19 20 21 22 23
2 24 25 26 27 28 29 30 31 32 33 34 35
3 36 37 38 39 40 41 42 43 44 45 46 47
4 48 49 50 51 52 53 54 55 56 57 58 59
Pandas希望对齐rhs一侧(毕竟,您要跟踪不同的索引), 因此您需要手动广播.这是关于此的问题: https://github.com/pydata/pandas/issues/7475
Pandas wants to align the rhs side (after all you are subtracing DIFFERENT indexes), so you need to manually broadcast this. Here is an issue about this: https://github.com/pydata/pandas/issues/7475
In [109]: df.loc[:,idx[:,:[2,3]]] = df.loc[:,idx[:,:,[2,3]]]-np.tile(df.loc[:,idx[:,:,1]].values,2)
Out[109]:
first A B
second a b a b
third 2 3 2 3 2 3 2 3
0 1 -1 -2 -4 7 5 4 2
1 1 -1 -2 -4 7 5 4 2
2 1 -1 -2 -4 7 5 4 2
3 1 -1 -2 -4 7 5 4 2
4 1 -1 -2 -4 7 5 4 2
这篇关于 pandas 具有多索引的高级横截面的文章就介绍到这了,希望我们推荐的答案对大家有所帮助,也希望大家多多支持IT屋!
查看全文