CODEWITHSUNDEEP
pythonscipy
Dr Manhattan
Dr Manhattan

Reputation: 14037

How does one calculate the probability of a multivariate Gaussian in Scipy?

How does one calculate the probability of a multivariate Gaussian at point x in Scipy?

I tried

    probability = multivariate_normal(x, mean, v)

where x, mean, v are all correctly sized vectors, but I get:

    probability = multivariate_normal(x, mean, v )
TypeError: __call__() takes at most 3 arguments (4 given)

Upvotes: 2

Views: 8391

Answers (1)

RickardSjogren
RickardSjogren

Reputation: 4238

First the error. When you call multivariate_normal you are actually calling __call__ of the multivariate_normal_gen class (source at row 555). Which since it is a method takes itself, self, as an argument which always is the first argument. When you then add three more arguments the total number of arguments is four.

Regarding your question, you evaluate the probability density function at given point(s) by calling pdf of multivariate_normal. Example:

from scipy.stats import multivariate_normal
import numpy as np
x = np.array([[1,2], [3,4]])
multivariate_normal.pdf(x, mean=[0, 1], cov=[5, 2])

Which prints out:

Out[44]: array([ 0.0354664 ,  0.00215671])

Upvotes: 7

Related Questions