CODEWITHSUNDEEP
pymcpymc3
Booley
Booley

Reputation: 859

PyMC3: How to create multiple random walks?

I want to create several random walk variables in PyMC3, or rather, stack several random walks into a single variable. I know I can create a single random walk of 100 steps like this:

with pm.Model() as model():
    z = pm.GaussianRandomWalk('z', mu=0, sd=1, shape=100)

But let's say I want to create 20 random walks, each of length 100. If I write this,

with pm.Model() as model():
    z = pm.GaussianRandomWalk('z', mu=0, sd=1, shape=(20,100))

Does that make every row a random walk, so I have 20 instances of 100 steps each? Or does it make every column a random walk, so I have 100 instances of 20 steps each?

Upvotes: 1

Views: 690

Answers (1)

Junpeng Lao
Junpeng Lao

Reputation: 424

You will have 100 instances of 20 steps each.

The time-series distributions in PyMC3 is not very rigour - it was designed with only 1D time series in mind, so do be careful when you are working with multiple time-series.

Upvotes: 0

Related Questions