Causal Inference: A Statistics Playground
A presentation from 2023–2024 Radcliffe fellow Judith Lok
At Radcliffe, Lok is writing “Causal Inference: A Statistics Playground,” a textbook designed for students and statisticians within and outside academia who work or intend to work in causal inference. Causal inference methods seek to address questions like “what would happen if” through data analysis. This textbook will primarily concentrate on non-randomized data, which are abundant. Estimating treatment effects from non-randomized data is challenging due to confounding by indication: when comparing treated and untreated individuals/units, differences arise not only from the treatment but also from pretreatment differences between the treated and untreated groups. Causal inference offers methods to overcome confounding by indication and other biases, allowing for the estimation of treatment effects from non-randomized data.