Closed rohanbabbar04 closed 3 months ago
Hi @aloctavodia ,
I think we can merge this PR, with the two tests and slight tweaks. I have opened a new issue #437 , I will be modifying the code and adding the tests for hierarchical models there itself(opening a new PR).
Better to add the test now. Just write separate tests for these 3 models. The other hierarchical model can wait.
normal_normal
with pm.Model() as model: x = pm.Normal("x", mu=0, sigma=10) z = pm.HalfNormal("z", 10) y = pm.Normal("y", x, z, observed=Y)
target = pz.Normal(mu=174, sigma=20)
norma_normal_shape
with pm.Model() as model: x = pm.Normal("x", mu=[0, 1], sigma=[10,10]) # This has shape 2 z = pm.HalfNormal("z", 10) y = pm.Normal("y", x[idx], z, observed=Y)
target = pz.Normal(mu=174, sigma=20)
hierarchical
with pm.Model(coords=coords) as model: μ_mu = pm.Normal('μ_mu', mu=0, sigma=10) μ_sd = pm.HalfNormal('μ_sd', 10) σ = pm.HalfNormal('σ', sigma=10)
μ = pm.Normal('μ', mu=μ_mu, sigma=μ_sd, dims="aa")
y = pm.Normal('y', mu=μ[idx], sigma=σ, observed=diff)
Closes #433
Description
Add tests for
ppe
.Return
new_priors
in ppe.Add
random_state
parameter to get a unique answer.