CODEWITHSUNDEEP
pythonnumpyscipy
Duna
Duna

Reputation: 735

Drawing from a continuous triangular distribution in python Scipy

How can I make draws and obtain the CDF at certain values x from a continuous triangular distribution with mode=0, lower limit=-1 and upper limit=1. I could not understand how to set the parameters. I want to get the equivalent of doing numpyp.random.triangular(left=-1, mode=0, right=1) in Scipy.

I attempted the following but I am not sure if that is what I am after.

scipy.stats.triang.cdf([-1,-0.5,0,0.5,1], c=0.5, loc=-1, scale=2) and obtained: array([ 0, 0.125, 0.5, 0.875, 1.]) which seems correct.

I could not why

scipy.stats.triang.expect(func=lambda x: x, c = 0.5, loc=0.5, scale=1)

is producing error message

_argcheck() missing 1 required positional argument: 'c'

though the c argument is provided.

Upvotes: 4

Views: 2823

Answers (1)

Warren Weckesser
Warren Weckesser

Reputation: 114921

It looks like you got the parameters of the CDF function correct. In general, if you have left, mode and right as used by numpy.random.triangular, you can convert those to the parameters of scipy.stats.triang using

c = (mode - left) / (right - left)
loc = left
scale = right - left

The arguments of the expect method can be confusing. You tried this:

scipy.stats.triang.expect(func=lambda x: x, c = 0.5, loc=0.5, scale=1)

The correct call is

In [91]: triang.expect(lambda x: x, (0.5,), loc=0.5, scale=1)
Out[91]: 1.0

The second argument is a tuple that holds the shape parameters, which in this case is the tuple (c,). There isn't a separate keyword argument for the shape parameters.

To draw samples from a triangular distribution with width 2 centered at 0, use the rvs method of scipy.triang, with c=0.5, loc=-1, and scale=2. For example, the following draws 10 samples:

In [96]: triang.rvs(c=0.5, loc=-1, scale=2, size=10)
Out[96]: 
array([-0.61654942,  0.03949263,  0.44191603, -0.76464285, -0.5474533 ,
        0.00343265,  0.222072  , -0.14161595,  0.46505966, -0.23557379])

Upvotes: 5

Related Questions