Prototyping panel data support for causal effect inference

[emrekiciman]

We are starting to prototype panel data support in causal effect inference. Creating this issue to track.

See also:

• panel-data branch: https://github.com/py-why/dowhy/tree/panel-data
• API proposal https://github.com/py-why/dowhy/wiki/API-proposal-for-v1
• discussion in Discord channel #panel-data https://discord.com/channels/818456847551168542/1015286112798916639

[Padarn]

Hey @emrekiciman thanks for starting this issue. I think it would be cool to get started with a prototype of the proposed API:

causalgraph = dowhy.panelanalysis.DiffInDiffGraph(outcomes = ['Y'],
                                            treatment_status = 'T',
                                            running_index = 't')

estimand = dowhy.identify_effect(causalgraph,
                                 action_node = ['T'],
                                 outcome_node = ['Y'],

I'm not totally clear on the state of the API update proposal, and if we have anything we should base a POC on?

I don't have access to edit the issue, but perhaps I could help break this down into some distinct tasks. I will probably start with the following tomorrow if no one tells me not to :-)

• [ ] Create poc version of DiffInDiffGraph creation of a causal Graph