Adding "unobserved"/ "latent" variables to DAG

[AnantChandra]

Hi there!

Went through this notebook & also ran a couple of experiments: https://www.pymc-marketing.io/en/stable/notebooks/mmm/mmm_causal_identification.html#the-real-challenge

Created this DAG:

causal_dag_str = """
digraph {
   *Spend columns*
   google_search_all_features_spend -> orders_kpi;
   ...
   
   *Control --> spend column*
   gqv_organic_google_search_impressions_control -> google_search_all_features_spend;

   *Control --> unobserved variable*
   gqv_organic_google_search_impressions_control -> brand_awareness;
   ...

   *Unobserved variable --> target column*
   brand_awareness -> orders_kpi;
}
"""

Why isn't my unobserved variable being reflected anywhere in the graphviz?

I ran another model after removing the following lines & got the exact same model output & fit as before.

   *Control --> unobserved variable*
   gqv_organic_google_search_impressions_control -> brand_awareness;
   ...

   *Unobserved variable --> target column*
   brand_awareness -> orders_kpi;

As I see it, my existing configuration of the "unobserved" variables isn't being captured.
Could someone please help me understand how to model multiple "unobserved"/ "latent" variables in a PyMC Marketing DAG 🙏

[AnantChandra]

@cetagostini would appreciate your help :)
Yet another DAG question

[AnantChandra]

Closing issue, as Juan replied on Slack.