Causal Inference: The Mixtape

Why I'm reading this:

Dense in places, but worth working through the section on confounding variables—it has direct implications for how we think about control testing in analytics. When we test a control by looking at outcomes, we're often implicitly assuming a causal relationship that may not hold. The discussion of selection bias is particularly relevant: populations that self-select into certain processes don't behave the same as random samples, and control testing analytics that ignore this can produce misleadingly clean results. This isn't a book about audit, but it's one of the more useful things I've read for thinking rigorously about audit evidence.