Confidence intervals for causal effects, using data collected in different experimental or environmental conditions. Hidden variables can be included in the model with a more experimental version.
