Handle broadcasting of parameters in Censored distributions

[ricardoV94]

When size is not specified, the base distribution is not resized and can broadcast with lower / upper, with repeated draws.

import aesara.tensor as at
import pymc as pm
rng = at.random.default_rng(123)
x = pm.Censored.dist(pm.Normal.dist(0), lower=[-1, -1, -1], upper=[1, 1, 1], rngs=(rng,))
x.eval()  # array([-0.98912135, -0.98912135, -0.98912135])

We need to do something like in:

pymc/pymc/distributions/timeseries.py

Lines 185 to 187 in 4702bd4

	# If not explicit, size is determined by the shapes of mu, sigma, and init
	bcast_shape = at.broadcast_arrays(mu, sigma, init)[0].shape
	init = resize_dist(init, bcast_shape)

[ghost]

Can you please be more specific? What should one to do fix this without changing base code?

[ricardoV94]

What do you mean without changing the base code?

[ricardoV94]

If you just want an immediate workaround, you can just make sure to pass size explicitly:

x = pm.Censored.dist(pm.Normal.dist(0), lower=[-1, -1, -1], upper=[1, 1, 1], size=3)