A Quickstart for Causal Analysis Decision-Making with DoWhy
Predict the Causal Effect from the intervention
Causal Analysis is an experimental analysis within the statistical field to establish cause and effect.
In the Data Analysis, we always have a concern with the Causal effect question, e.g. Is paracetamol 500 mg cure headache, or would someone buy my 5-years old Laptop?. This kind of question often analyzed from the statistical perspective with respect to the available data.
While the golden boy to know the causal effect is the A/B Testing, what if the testing is not possible for some reasons, e.g. time-constraint, cost, or just simply no data. This is where we could applying the Causal Analysis to estimate the effect of the intervention (Feature) on the outcome.
Causal analysis is inherently different compared to the prediction that came from Machine Learning modeling. While we could try to predict the results via a model that learns the data pattern, but we never knew what happens outside of the data dimension.
Just imagine this, you might have an exam tomorrow and decided to study for two hours straight. The outcome is your exam score with the intervention of two hours of study, but what if you only study for one hour? would there any effect? We cannot turn back…
