采用python的scipy库完成常用的假设检验, 配合pandas库非常好用

正态性检验

检验数据样本是否具有高斯分布。

from scipy.stats import shapiro

data = [21,12,12,23,19,13,20,17,14,19]

stat,p = shapiro(data)

print("stat为：%f" %stat,"p值为：%f" %p)

皮尔逊相关性检验

检查两个样本是否相关的统计检验

from scipy.stats import pearsonr

data1 = [21,12,12,23,19,13,20,17,14,19]

data2 = [12,11,8,9,10,15,16,17,10,16]

corr,p = pearsonr(data1,data2)

print("corr为：%f" %corr,"p值为：%f" %p)

卡方检验

检验两个分类变量的独立性

from scipy.stats import chi2_contingency

data1 = [21,12,12,23,19,13,20,17,14,19]

data2 = [12,11,8,9,10,15,16,17,10,16]

stat,p,dof,expected = chi2_contingency(data1,data2)

print("stat为：%f" %stat,"p值为：%f" %p)

T检验

检验两个独立样本的均值是否存在显著差异

from scipy.stats import ttest_ind

data1 = [21,12,12,23,19,13,20,17,14,19]

data2 = [12,11,8,9,10,15,16,17,10,16]

stat,p = ttest_ind(data1,data2)

print("stat为：%f" %stat,"p值为：%f" %p)

配对T检验

检验两个配对样本的均值是否有显著差异

from scipy.stats import ttest_rel

data1 = [21,12,12,23,19,13,20,17,14,19]

data2 = [12,11,8,9,10,15,16,17,10,16]

stat,p = ttest_rel(data1,data2)

print("stat为：%f" %stat,"p值为：%f" %p)

方差分析

测试两个或两个以上独立样本的均值是否存在显著差异

from scipy.stats import f_oneway

data1 = [21,12,12,23,19,13,20,17,14,19]

data2 = [12,11,8,9,10,15,16,17,10,16]

stat,p = f_oneway(data1,data2)

print("stat为：%f" %stat,"p值为：%f" %p)

Mann-WhitneyU检验

检验两个独立样本的分布是否相等. 对应参数检验里面的t检验.

from scipy.stats import mannwhitneyu

data1 = [21,12,12,23,19,13,20,17,14,19]

data2 = [12,11,8,9,10,15,16,17,10,16]

stat,p = mannwhitneyu(data1,data2)

print("stat为：%f" %stat,"p值为：%f" %p)

Wilcoxon符号秩检验

检验两个配对样本的分布是否均等. 对应参数检验里面的配对t检验.

from scipy.stats import wilcoxon

data1 = [21,12,12,23,19,13,20,17,14,19]

data2 = [12,11,8,9,10,15,16,17,10,16]

stat,p = wilcoxon(data1,data2)

print("stat为：%f" %stat,"p值为：%f" %p)

Kruskal-wallis

H检验.

检验两个或多个独立样本的分布是否相等. 对应参数检验里面的方差分析.

from scipy.stats import kruskal

data1 = [21,12,12,23,19,13,20,17,14,19]

data2 = [12,11,8,9,10,15,16,17,10,16]

stat,p = kruskal(data1,data2)

print("stat为：%f" %stat,"p值为：%f" %p)