how does it compare to causalinference library about dowhy HOT 2 CLOSED

Thanks for your interest. The library laurencium/causalinferece also implements causal inference methods, but it focuses only on one step in causal analysis: estimation. In building DoWhy, therefore, we decided to have support for all the steps required for a causal analysis:

• model (make assumptions),
• identify (find what to estimate given the assumptions),
• estimate
• refute (sensitivity and robustness checks).

I think the biggest difference is the last step: DoWhy provides refute methods to test underlying causal assumptions. This is often missed in causal inference libraries that only deal with estimation, as the user is expected to know beforehand what is the right causal target estimand, and which covariates to include. DoWhy also allows the user to specify your asssumptions explicitly in code, so that they can be reasoned about and tested if possible.

Finally, in terms of estimation methods, DoWhy supports both backdoor-based and instrumental variable methods, while laurencium/causalinference library supports only backdoor-based methods.

from dowhy.

@amit-sharma Sounds awesome! Thank you for taking time to answer my question. I'll definitely delve more into the inner workings in the next couple of days. I'm closing this issue in hope to keep things clean since your answer is really thorough.

from dowhy.