Fitting a Weibull distribution using Scipy

Question

I am trying to recreate maximum likelihood distribution fitting, I can already do this in Matlab and R, but now I want to use scipy. In particular, I would like to estimate

Answer 1

I was curious about your question and, despite this is not an answer, it compares the Matlab result with your result and with the result using leastsq, which showed the best correlation with the given data:

The code is as follows:

import scipy.stats as s
import numpy as np
import matplotlib.pyplot as plt
import numpy.random as mtrand
from scipy.integrate import quad
from scipy.optimize import leastsq

## my distribution (Inverse Normal with shape parameter mu=1.0)
def weib(x,n,a):
    return (a / n) * (x / n)**(a-1) * np.exp(-(x/n)**a)

def residuals(p,x,y):
    integral = quad( weib, 0, 16, args=(p[0],p[1]) )[0]
    penalization = abs(1.-integral)*100000
    return y - weib(x, p[0],p[1]) + penalization

#
data = np.loadtxt("stack_data.csv")


x = np.linspace(data.min(), data.max(), 100)
n, bins, patches = plt.hist(data,bins=x, normed=True)
binsm = (bins[1:]+bins[:-1])/2

popt, pcov = leastsq(func=residuals, x0=(1.,1.), args=(binsm,n))

loc, scale = 1.58463497, 5.93030013
plt.plot(binsm,n)
plt.plot(x, weib(x, loc, scale),
         label='weib matlab, loc=%1.3f, scale=%1.3f' % (loc, scale), lw=4.)
loc, scale = s.exponweib.fit_loc_scale(data, 1, 1)
plt.plot(x, weib(x, loc, scale),
         label='weib stack, loc=%1.3f, scale=%1.3f' % (loc, scale), lw=4.)
plt.plot(x, weib(x,*popt),
         label='weib leastsq, loc=%1.3f, scale=%1.3f' % tuple(popt), lw=4.)

plt.legend(loc='upper right')
plt.show()