PyMC v5.10.3 prediction stuff

Questions v5
9

Hi Ricardo. Thank you for your answer
Now it seems there’s a shape issue…
…maybe I’m getting lost in a glass of water…
I’ll present the entire code to you below.

control_vars = ['c1','c2', 'c3', 'c4', 'c5' ] 
target = df_scaled['y'].to_numpy()
df_controls  = df_scaled[control_vars]
n_obs, n_controls = df_controls.shape

coords = {'controls':control_vars, 
          'all_vars':control_vars}

with pm.Model(coords=coords) as model:
    X = pm.MutableData('control_data', df_controls.values)
    y = pm.MutableData('targets', df_scaled['y'].values.squeeze())

    n_obs = X.shape[0]
    
    contributions = []
    
    control_betas = pm.Normal('control_beta', sigma = 2, dims=['controls'])
    
    for w in range(n_controls):
        x = X[:,w]*control_betas[w]
        contributions.append(x)
         
    mu = pm.Deterministic("contributions", tt.stack(contributions).T, dims=['controls'])
    sigma = pm.HalfNormal('sigma', sigma=1)
    
    y_hat = pm.Normal("y_hat", mu=mu.sum(axis=-1), sigma=sigma, observed=target, shape=X.shape[0])
    
with model:
    idata = pm.sample(idata_kwargs={'dims':{'contributions':[None, 'controls']}})

And up to this point, everything is fine, then there’s the out-of-sample part.

x_test = df_test_controls.values
with model:
    pm.set_data({"control_data":x_test})
    idata.extend(pm.sample_posterior_predictive(idata, var_names=["y_hat", "contributions"]))

But I obtain the following error:

ValueError: conflicting sizes for dimension ‘controls’: length 110 on the data but length 5 on coordinate ‘controls’

the shapes are:
df_test_controls.values = Array of float64 (110,5)
df_controls.values = Array of float64 (110,5)
coords = Dict (5)

It seems there is a mismatch in dimensions. Perhaps a naming issue?
Thank you