# Everyday Causal Inference > A practical, code-first guide to causal inference for data scientists, product managers, and analysts. The book teaches the craft of causal inference — not just the math — with a focus on reasoning, critical thinking, and translating statistical findings into business decisions. All examples use R and Python with real-world scenarios from e-commerce, fintech, and tech platforms. ## Author - [Robson Tigre](https://www.robsontigre.com), Ph.D. in Economics. Industry practitioner in fintech and experimentation platforms. ## Chapters ### Part I: Foundations - [Data, Models, and Causality](https://www.everydaycausal.com/data-models-causality.html): Explains why machine learning prediction and causal inference answer fundamentally different questions. Covers OLS regression as a causal tool, the distinction between population and sample, and the estimand-estimator-estimate framework that structures every causal analysis. Answers: What is the difference between prediction and causal inference? How do I interpret regression coefficients causally? - [Biases and Causal Frameworks](https://www.everydaycausal.com/causal-frameworks.html): Breaks down the anatomy of omitted variable bias with full mathematical intuition, explains selection bias, and introduces the potential outcomes framework. Defines and distinguishes ATE, ATT, ITT, LATE, CATE, and ITE with practical examples showing when each estimand matters. Answers: What is omitted variable bias and why does it matter? What is the difference between ATE and ATT? - [Planning Your Analysis](https://www.everydaycausal.com/planning-your-analysis.html): How to formulate a causal question before touching data, avoid HARKing (hypothesizing after results are known), and design a credible causal study. Includes a worked example of planning an A/B test from question to analysis specification. Answers: How do I write a pre-analysis plan? What makes a causal question well-defined? ### Part II: Estimating impacts - [Designing Experiments (RCTs)](https://www.everydaycausal.com/designing-experiments.html): Covers the practical mechanics of running A/B tests: SUTVA assumptions, internal and external validity, choosing metrics (OEC, driver, guardrail), randomization schemes, pre-flight sanity checks, and experiment monitoring. Answers: What is SUTVA and when does it fail? How do I choose the right metric for my experiment? - [Power Analysis](https://www.everydaycausal.com/power-analysis.html): Explains hypothesis testing, Type I and Type II errors, minimum detectable effect (MDE), and sample size calculations with intuitive explanations and code. Shows how to compute required sample sizes and interpret power curves. Answers: How many users do I need for my A/B test? What is statistical power and why does it matter? - [Causal Assumptions (DAGs)](https://www.everydaycausal.com/causal-assumptions.html): Teaches how to use directed acyclic graphs to map causal relationships and identify what to control for. Covers frontdoor and backdoor paths, confounders, mediators, colliders, and the backdoor criterion — all with visual diagrams and practical examples. Answers: How do I use a DAG to decide which variables to control for? What is a collider and why should I not control for it? - [Instrumental Variables](https://www.everydaycausal.com/instrumental-variables.html): Explains how to estimate causal effects when treatment assignment is not random, using instruments that affect the outcome only through the treatment. Covers two-stage least squares, the exclusion restriction, monotonicity, compliance types (always-takers, never-takers, compliers), weak instruments, and the LATE framework. Answers: What is an instrumental variable and when do I need one? What is the difference between ITT and LATE? - [Regression Discontinuity Design](https://www.everydaycausal.com/regression-discontinuity.html): Shows how to use cutoff-based rules as natural experiments. Covers sharp and fuzzy RDD, running variable selection, bandwidth choice, polynomial order, and validation tests (density, covariate balance, placebo cutoffs). Includes R and Python implementations with rdrobust. Answers: What is a regression discontinuity design? How do I choose the bandwidth for RDD? - [TWFE Difference-in-Differences](https://www.everydaycausal.com/twfe-did.html): Estimates treatment effects by comparing changes over time between treated and control groups. Covers the parallel trends assumption, two-way fixed effects specification, event study plots, and pre-trend testing. Answers: What is the parallel trends assumption? How do I implement difference-in-differences in R or Python? - [Staggered DiD](https://www.everydaycausal.com/staggered-did.html): Addresses the problems that arise when treatment is adopted at different times by different units. Explains why standard TWFE fails with staggered timing (negative weighting, heterogeneity bias) and introduces modern estimators like Callaway and Sant'Anna that produce valid estimates. Answers: Why does standard DiD fail with staggered treatment? How do I use the Callaway and Sant'Anna estimator? - [Time Series Causal Inference](https://www.everydaycausal.com/time-series.html): Measures causal impact when there is no control group, using pre-intervention data to build synthetic counterfactuals. Covers Google's CausalImpact (Bayesian structural time series), CausalArima, the role of auxiliary series, and pre/post intervention analysis. Answers: How do I measure causal impact without a control group? What is CausalImpact and how does it work? - [Heterogeneous Effects](https://www.everydaycausal.com/heterogeneous-effects.html): Explores who benefits most (or least) from a treatment. Covers conditional average treatment effects (CATE), causal forests, and meta-learners (S-learner, T-learner, X-learner) for estimating treatment effect heterogeneity. Answers: How do I estimate treatment effects for different subgroups? What is a causal forest? ### Part III: Useful, credible estimates, and next steps - [Falsification and Robustness](https://www.everydaycausal.com/falsification.html): Stress-tests a causal estimate before someone else does. Covers the placebo family (placebo-in-time, negative-control outcomes and treatments, placebo units), balance and overlap diagnostics, the conditional independencies a DAG implies, robustness and specification grids, and sensitivity analysis for unobserved confounding (partial R-squared, robustness values, benchmarking). Answers: How do I know my causal estimate is not driven by confounding? What is a placebo test and how do I run one? How sensitive is my result to a confounder I cannot observe? - [Business Translations](https://www.everydaycausal.com/business-translations.html): Bridges the gap between statistical estimates and business decisions. Covers how to translate causal effects into monetary value, compute ROI/NPV/IRR, estimate payback periods, model CAC-to-LTV ratios, account for carryover and decay effects, build scaling forecasts with diminishing returns, and make go/no-go investment decisions. Answers: How do I convert a treatment effect into ROI? How do I present causal inference results to a CFO? - [Other Approaches](https://www.everydaycausal.com/other-approaches.html): Maps what lies beyond this book's core toolkit. Covers structural causal models and do-calculus, causal discovery, causal machine learning (double machine learning, meta-learners), Bayesian causal inference, sequential and time-varying treatments, interference and spillovers, policy learning, and mechanisms and mediation. Answers: What do I do when my treatment changes over time? How do I handle spillovers between users? Whom should I treat, not just what is the effect? ## Resources - [Glossary](https://www.everydaycausal.com/glossary.html): Definitions of causal inference terms including ATE, ATT, CATE, collider, confounder, DAG, instrumental variable, LATE, parallel trends, SUTVA, and more. - [GitHub Repository](https://github.com/RobsonTigre/everyday-ci): Full source code, datasets, and reproducible examples for all chapters. - [How to Use This Book](https://www.everydaycausal.com/how-to-use-it.html): Reading guide with recommended paths based on background and goals.