如何绘制基于数据列的Mann检验(Wilcoxon秩和检验)？

Question

我想绘制基于subtype列的Mann检验(Wilcoxon秩和检验)，该列包含normal、KIRP、KIRC、KICH数据。

我刚开始使用Python进行统计。

import pandas as pd
import plotly.express as px
from scipy.stats import mannwhitneyu


normal = df.iloc[:,7:-2][df['subtype'] == "normal"]
kirp = df.iloc[:,7:-2][df['subtype'] == "KIRP"]
kirc = df.iloc[:,7:-2][df['subtype'] == "KIRC"]
kich = df.iloc[:,7:-2][df['subtype'] == "KICH"]

nnormal, nkirp, nkirc, nkich = len(normal), len(kirp), len(kirc), len(kich)

U1, p = mannwhitneyu(normal, kirp, kirc, kich, method="exact")

回溯：

---------------------------------------------------------------------------
TypeError                                 Traceback (most recent call last)
/tmp/ipykernel_27/1709062782.py in <module>
----> 1 U1, p = mannwhitneyu(normal, kirp, kirc, kich, method="exact")

/opt/conda/lib/python3.7/site-packages/scipy/stats/_mannwhitneyu.py in mannwhitneyu(x, y, use_continuity, alternative, axis, method)
    389 
    390     x, y, use_continuity, alternative, axis_int, method = (
--> 391         _mwu_input_validation(x, y, use_continuity, alternative, axis, method))
    392 
    393     x, y, xy = _broadcast_concatenate(x, y, axis)

/opt/conda/lib/python3.7/site-packages/scipy/stats/_mannwhitneyu.py in _mwu_input_validation(x, y, use_continuity, alternative, axis, method)
    136 
    137     bools = {True, False}
--> 138     if use_continuity not in bools:
    139         raise ValueError(f'`use_continuity` must be one of {bools}.')
    140 

TypeError: unhashable type: 'DataFrame'

下游代码：

U2 = nnormal*nkirp*nkirc*nkich - U1
_, pnorm = mannwhitneyu(normal, kirp, kirc, kich, method="asymptotic")

# Plot
fig = px.box(_, y="pnorm")
fig.show()

期望输出示例：

​

Answer 1

mannwhitneyu方法接受两个样本数组。您会得到错误，因为函数的第三个参数(use_continuity)可以是True或False。如果要将kirp、kirc和kich与正常样本进行比较，可以这样做：

u_kirp, p_kirp = mannwhitneyu(normal, kirp, method="exact")
u_kirc, p_kirc = mannwhitneyu(normal, kirc, method="exact")
u_kich, p_kich = mannwhitneyu(normal, kich, method="exact")