python - Python scipy chisquare 返回与 R chisquare 不同的值

Question

我正在尝试使用scipy.stats.chisquare. 我建立了一个玩具示例：

In [1]: import scipy.stats as sps

In [2]: import numpy as np

In [3]: sps.chisquare(np.array([38,27,23,17,11,4]), np.array([98, 100, 80, 85,60,23]))
Out[11]: (240.74951271813072, 5.302429887719704e-50)

返回中的相同示例R：

> chisq.test(matrix(c(38,27,23,17,11,4,98,100,80,85,60,23), ncol=2))

Pearson's Chi-squared test

data:  matrix(c(38, 27, 23, 17, 11, 4, 98, 100, 80, 85, 60, 23), ncol = 2)
X-squared = 7.0762, df = 5, p-value = 0.215

我究竟做错了什么？

谢谢

Answer 1

对于此chisq.test调用，python 等效项是chi2_contingency：

此函数计算观察到的列联表中观察到的频率独立性的假设检验的卡方统计量和 p 值。

>>> arr = np.array([38,27,23,17,11,4,98,100,80,85,60,23]).reshape(2,-1)
>>> arr
array([[ 38,  27,  23,  17,  11,   4],
       [ 98, 100,  80,  85,  60,  23]])
>>> chi2, p, dof, expected = scipy.stats.chi2_contingency(arr)
>>> chi2, p, dof
(7.0762165124844367, 0.21503342516989818, 5)