数据分布假设检验的p值适合python scipy

如何解决数据分布假设检验的p值适合python scipy

我正在尝试使用假设检验来获得数据分布。 据说给定的数据遵循Gamma分布（使用JS散度和MLE测试获得）。但是，p检验不能揭示相同的事实。任何人都可以检查此代码，并让我知道代码中的错误吗？预先谢谢你！

import numpy as np
import matplotlib.pyplot as plt
import scipy.stats as sts



data=np.asarray([113,116,113,115,123,126,135,128,141,144,137,142,146,149,152,153,119,118,117,120,129,131,127,143,156,158,121,124,130,133,125,107,111,97,102,109,112,93,92,95,100,103,110,105,96,94,99,101,104,91,108,89,88,93])



dist1 = [sts.norm,sts.uniform,sts.expon,sts.logistic,sts.lognorm,sts.gamma]
mles1 = []
#### distribution fit using MLE
for distribution in dist1:
    pars = distribution.fit(data)
    mle = distribution.nnlf(pars,data)
    mles1.append(mle)
print("Results of MLE")
results1 = [(distribution.name,mle) for distribution,mle in zip(dist1,mles1)]
print(results1)

data = (data - data.min())/(data.max() - data.min())  # Normalize data
stat,p = sts.shapiro(data)
print('Shapiro --- stat=%.3f,p=%.3f' % (stat,p))
if p > 0.05:
    print('Probably Gaussian')
else:
    print('Probably not Gaussian')
    
args1 = sts.norm.fit(data)
stat,p = sts.kstest(data,'norm',args1)
print("KS test for Normal distribution p value ",p)

args2 = sts.gamma.fit(data)
print("KS test for Gussian distribution p value",sts.kstest(data,'gamma',args2))



## pdf fit using histogram
entries,bins,patches = plt.hist(data.ravel(),50,facecolor = "plum",density=True)
dist = getattr(sts,"gamma")
param = dist.fit(bins)
pdf_fitted = dist.pdf(bins,*param[:-2],loc=param[-2],scale=param[-1])


stat,p = sts.f_oneway(bins,pdf_fitted)
print('stat=%.3f,p))
if p > 0.05:
    print('Probably gamma distribution')
else:
    print('Probably not gamma distributions')
    

stat,p = sts.ansari(bins,p))
if p > 0.05:
    print("Probably same scale parameters")
else:
    print("Probably different scale parameters")

上面的代码给出如下所示的输出。如圆圈所示，MLE的最小值是Gamma分布的。但是，KS测试的p值太低。