CODEWITHSUNDEEP
pythonnumpyscipy
bcf
bcf

Reputation: 2134

Multivariate Normal pdf in Scipy

Trying to evaluate scipy's multivariate_normal.pdf function, but keep getting errors. MWE:

import numpy as np
from scipy.stats import multivariate_normal as mvnorm

x = np.random.rand(5)
mvnorm.pdf(x)

gives

TypeError: pdf() takes at least 4 arguments (2 given)

The docs say both the mean and cov arguments are optional, and that the last axis of x labels the components. Since x.shape = (4L,), it seems like all is kosher. I am expecting a single number as output.

Upvotes: 4

Views: 6452

Answers (1)

dot.Py
dot.Py

Reputation: 5157

It looks like these parameters aren't optional.

If I pass the default values for mean and cov like:

import numpy as np
from scipy.stats import multivariate_normal as mvnorm

x = np.random.rand(5)
mvnorm.pdf(x, mean=0, cov=1)

I get the following output:

array([ 0.35082878,  0.27012396,  0.26986049,  0.39887847,  0.36116341])

While using:

import numpy as np
from scipy.stats import multivariate_normal as mvnorm

x = np.random.rand(5)
mvnorm.pdf(x)

gives me the same error:

TypeError: pdf() takes at least 4 arguments (2 given)

Upvotes: 3

Related Questions