Vaishnavi K This model was exclusively learned as a predictive model. Hence, a distinction between confounders and non-confounders is a non-issue. We would only have to make that distinction if we had the objective of performing causal inference, which was not the case here.
Stefan Conrady If we had to perform causal inference on this data, what should be my confounders/Non-confounders. Few examples might help. Also, would the result change when we perform causal inference on this data.
Vaishnavi K In this particular case, one could not simply switch to causal inference. For instance, we know this data is a subset of a much larger dataset. Thus, the assumption of "no unobserved confounders" cannot be justified at all. So, it's a bad example for causal inference, I'm afraid.