关于numpy和pandas中std()函数的区别

今天在做实验时，发现pandas中std（）函数计算出来的标准差与手工计算出来的值存在差异，怀疑之下，经查阅文档，发现pandas的std（）与numpy的std（）存在差异，实验流程如下；

import pandas as pd
import numpy as np
# df = pd.DataFrame([['a','b','c','d','e','f'],[1,2,3,4,5,6]],index=['one','two'],columns=['A','B','C','D','E','F'])
df = pd.DataFrame([[1,2,3,4,5,6],[6,4,1,2,3,5],[2,43,2,2,7,9]],index=['one','two','three'],columns=['A','B','C','D','E','F'])
print(df)
# print(df.count(axis=1))
# print(df.sum(1))
# print(df.mean())
# print(df.mode(axis=0))
# print(df.mode(axis=1))
print(df.std())
# print(df.describe())
# print(np.sqrt())
l1 = np.array([1,2,3])
print(l1.std())

从结果中我们可以看到，pandas计算出来的std与numpy计算出来的不一致。

根据手工计算，1,2,3的均值为2，方差为(（1-2）^2+（2-2）^2+（3-2）^2)/3=2/3=0.666667，std肯定不会为1，所以numpy计算的应该是与我们知道情况一致；

但pandas的又是怎么回事呢，经计算，出现上述情况是pandas在计算过程中使用了index的标签所导致的，按照pandas的index标签，第一行的标签为0，第3行的标签为2，方差为(（1-2）^2+（2-2）^2+（3-2）^2)/2 = 1，所以std为1。

要想正常计算pandas的std，需要建ddof设置为0即可；pandas的ddof默认为1；

import pandas as pd
import numpy as np
# df = pd.DataFrame([['a','b','c','d','e','f'],[1,2,3,4,5,6]],index=['one','two'],columns=['A','B','C','D','E','F'])
df = pd.DataFrame([[1,2,3,4,5,6],[6,4,1,2,3,5],[2,43,2,2,7,9]],index=['one','two','three'],columns=['A','B','C','D','E','F'])
print(df)
# print(df.count(axis=1))
# print(df.sum(1))
# print(df.mean())
# print(df.mode(axis=0))
# print(df.mode(axis=1))
print(df.std(ddof=0))
# print(df.describe())
# print(np.sqrt())
l1 = np.array([1,2,3])
print(l1.std())

需要注意的是在pandas中的describe（）函数的统计中也有std信息，这里计算出来的std与未调整之前的std保持一致，且没有ddof设置，无法调整该值。

import pandas as pd
import numpy as np
# df = pd.DataFrame([['a','b','c','d','e','f'],[1,2,3,4,5,6]],index=['one','two'],columns=['A','B','C','D','E','F'])
df = pd.DataFrame([[1,2,3,4,5,6],[6,4,1,2,3,5],[2,43,2,2,7,9]],index=['one','two','three'],columns=['A','B','C','D','E','F'])
print(df)
# print(df.count(axis=1))
# print(df.sum(1))
# print(df.mean())
# print(df.mode(axis=0))
# print(df.mode(axis=1))
print(df.std(ddof=0))
print(df.describe())
# print(np.sqrt())
l1 = np.array([1,2,3])
print(l1.std())

       A   B  C  D  E  F
one    1   2  3  4  5  6
two    6   4  1  2  3  5
three  2  43  2  2  7  9
A     2.160247
B    18.873850
C     0.816497
D     0.942809
E     1.632993
F     1.699673
dtype: float64
              A          B    C         D    E         F
count  3.000000   3.000000  3.0  3.000000  3.0  3.000000
mean   3.000000  16.333333  2.0  2.666667  5.0  6.666667
std    2.645751  23.115651  1.0  1.154701  2.0  2.081666
min    1.000000   2.000000  1.0  2.000000  3.0  5.000000
25%    1.500000   3.000000  1.5  2.000000  4.0  5.500000
50%    2.000000   4.000000  2.0  2.000000  5.0  6.000000
75%    4.000000  23.500000  2.5  3.000000  6.0  7.500000
max    6.000000  43.000000  3.0  4.000000  7.0  9.000000
0.816496580927726

Process finished with exit code 0