在堆积矩阵上运行kendall-tau回归的更快方法？

Question

我有一些类似于时间序列的矩阵，我想通过每一列和每一行的组合来运行kendall tau回归。我可以使用以下代码执行此操作：

import numpy as np
import pandas as pd
import scipy as sp
import random

#make some random data
test = []
for i in range(50):
    m = np.random.random((50, 50))
    test.append(m)


#copy a matrix to repopulate with the tau value  
coef = test[0].copy()

#copy to repopulate with a p-value
p = test[0].copy()

stacked = np.dstack(test)

#loop through each pixel and run the trend test, output two matrix, one of slope and one of p-value
for r in range(test[0].shape[0]):
    for c in range(test[0].shape[1]):

        pixel = list(stacked[r: r+1, c:c+1, :].flatten()) 

        df = pd.DataFrame({"Values": pixel})

        #reset the index to use as a independent variable
        df = df.reset_index(drop = True)

        #run the regression
        tau, p_value = sp.stats.kendalltau(df.index, df.Values)

        #update the copied matrices
        coef[r: r+1, c:c+1] = tau
        p[r: r+1, c:c+1] = p_value

我想知道是否有一个更有效的方法来完成这项任务，虽然循环通过堆叠矩阵中的每个元素似乎是不正确的方式。也许有办法把它矢量化？你知道吗