Search code examples
pythonmatplotlibscipyprobability

How to properly plot the pdf of a beta function in scipy.stats

I am trying to fit a beta distribution to some data, and then plot how well the beta distribution fits the data. But the output looks really weird and incorrect.

import scipy.stats as stats 
import matplotlib.pyplot as plt

x = np.array([0.9999999 , 0.9602287 , 0.8823198 , 0.83825594, 0.92847216,
       0.9632976 , 0.90275735, 0.8383094 , 0.9826664 , 0.9141795 ,
       0.88799196, 0.9272752 , 0.94456017, 0.90466917, 0.8905505 ,
       0.95424247, 0.781545  , 0.9489085 , 0.9578988 , 0.8644015 ])
beta_params = stats.beta.fit(x)
print(beta_params)
#(3.243900357315478, 1.5909897101396109, 0.7270083219563888, 0.27811444901271615
 
beta_pdf = stats.beta.pdf(x, beta_params[0], beta_params[1], beta_params[2], beta_params[3])

print(beta_pdf)
#[2.70181543 6.8442073  4.98204632 2.82445508 6.76055614 6.75910611
 #5.90419012 2.82696622 5.58521916 6.34096675 5.2508072  6.73212694
 #6.98854653 5.98225724 5.36937625 6.9519977  0.67812362 6.99116729
 #6.89484982 4.10113147]

plt.plot(x, beta_pdf)

enter image description here

Solution

  • I'm not a statistician, but looking at your code I see that x is unordered.

    Does sorting x before fit helps you?

    x = np.sort(x)
beta_params = stats.beta.fit(x)

    Doing so, you'd get this:

    enter image description here