Compact plot to plot both Traces in single plot

[sudarsan3010]

Dear community,

I am new to the statistical plotting and data processing. I was trying with a simple timeline.

compact=True does not work for me. I am giving my code below.

# Bayesian Hierarchical Model with PyMC3
with pm.Model() as model:
    # Priors for coefficients
    alpha = pm.Normal("alpha", mu=0, sigma=10)  # Intercept
    beta_rain = pm.Normal("beta_rain", mu=0, sigma=10)  # Coefficient for rainfall
    beta_et = pm.Normal("beta_et", mu=0, sigma=10)  # Coefficient for ET
    sigma = pm.HalfNormal("sigma", sigma=1)  # Noise term

    # Expected recharge
    mu = alpha + beta_rain * rainfall_standardized + beta_et * et_standardized

    # Likelihood
    recharge_obs = pm.Normal("recharge_obs", mu=mu, sigma=sigma, observed=recharge_standardized)

    # Sampling
    trace = pm.sample(1000, return_inferencedata=True)

Trial plot 1:

az.plot_trace(trace, figsize=(9, 8), compact=True) 
plt.tight_layout()  # Ensure the layout does not overlap
plt.show()

Trial plot 2:
`

fig, axes = plt.subplots(2, 2, figsize=(12, 10))  
az.plot_trace(trace, var_names=["beta_rain", "beta_et"], axes=axes, legend=True, compact=True)

Expected results:
My two variables's traces should be plotted together in different colour

[aloctavodia]

Hi @sudarsan3010

What compact=True does is to plot multidimensional variables in a single plot. Instead, you have two different unidimensional variables.

[OriolAbril]

If variables are multidimensional or unidimensional is defined when creating the model, in your case having both betas as a multidimensional variable would mean using a batched normal when defining them, for example:

with pm.Model(coords={"coefficient": ["rainfall", "ET"]}) as model:  # add labeled coordinate values
    alpha = pm.Normal("alpha", mu=0, sigma=10) 
    beta = pm.Normal("beta_rain", mu=0, sigma=10, dims="coefficient") # use batched normal via dims argument
    sigma = pm.HalfNormal("sigma", sigma=1) 

    # Expected recharge
    mu = alpha + beta[0] * rainfall_standardized + beta[1] * et_standardized  # use beta accordingly

    # Likelihood
    recharge_obs = pm.Normal("recharge_obs", mu=mu, sigma=sigma, observed=recharge_standardized)

    # Sampling
    trace = pm.sample(1000, return_inferencedata=True)