Baysian hirerchical Linear Regression(Partial Pooling) model using PYMC

Questions version agnostic
10

i updated code as


# Define the model
with pm.Model() as hierarchical_model:
    #Define the data
    Location = pm.Data('LocationDesc_encoded', training_data['LocationDesc_encoded'],mutable=True)
    Obesity = pm.Data('obesity_Prevalence', training_data['obesity_Prevalence'],mutable=True)
    Data_val_py = pm.Data('data_value_py', training_data['data_value_py'],mutable=True)
    Avg_Temp = pm.Data('Avg Temp(°F)', training_data['Avg Temp(°F)'],mutable=True)
    AQI = pm.Data('AQI', training_data['AQI'],mutable=True)


    
    # Hyperpriors for group nodes
    mu_a = pm.Normal('mu_a', mu=0, sigma=0.5)
    sigma_a = pm.HalfCauchy('sigma_a', beta=1)

    # Priors for individual intercepts
    a = pm.Normal('a', mu=mu_a, sigma=sigma_a, shape=len(training_data['LocationDesc_encoded'].unique()))

    # Hyperpriors for group slopes
    mu_b = pm.Normal('mu_b', mu=0, sigma=1)
    sigma_b = pm.HalfCauchy('sigma_b', beta=1)

    # Priors for individual slopes
    b = pm.Normal('b', mu=mu_b, sigma=sigma_b, shape=(len(training_data['LocationDesc_encoded'].unique()), 4))

    # Model error
    sigma = pm.HalfCauchy('sigma', beta=1)

    # Expected value
    mu = a[Location] + \
         b[Location, 0] * Obesity + \
         b[Location, 1] * Data_val_py + \
         b[Location, 2] * Avg_Temp + \
         b[Location, 3] * AQI

    # Likelihood
    Y_obs = pm.Normal('Y_obs', mu=mu, sigma=sigma, observed=training_data['Data_Value'])


with hierarchical_model:
    idata = pm.sample(100, chains=2, target_accept=0.9,random_seed = rng)

with  hierarchical_model:
  pm.set_data({
        'LocationDesc_encoded': test_data['LocationDesc_encoded'],
        'obesity_Prevalence': test_data['obesity_Prevalence'],
        'data_value_py': test_data['data_value_py'],
        'Avg Temp(°F)': test_data['Avg Temp(°F)'],
        'AQI': test_data['AQI']
    })
  posterior_predictive = pm.sample_posterior_predictive(idata, random_seed=rng)

Now it is giving shape mismatch error because training data is (392 x 8) and test data is of (97 x 8).

To resolve this I tried

with  hierarchical_model:
  pm.Model.set_dim('LocationDesc_encoded', len(set(test_data['LocationDesc_encoded'])))

but this is giving error as

TypeError: Model.set_dim() missing 1 required positional argument: 'new_length'

Please help to resolve issue!