Causal Inference on Observational Data: It's All About the Assumptions
Dataiku
JULY 9, 2021
Previously , we showed that uplift modeling, a causal inference success story for businesses, can outperform more conventional churn models. As with any causal inference application, it relied on crucial assumptions about the data to correctly identify the causal effect. While we brushed those assumptions aside, contenting ourselves with the assertion that they hold whenever the treatment variable was randomized, we will present and examine the two fundamental assumptions of ignorability and pos
Let's personalize your content