The Data Analyst’s Guide to Cause and Effect

Theiss Bendixen is a PhD, quantitative consultant, and independent researcher. To date, he has written two popular science books, a co-editedvolume, as well as more than 40 academic works, including tutorials, quantitative empirical papers, and technical commentaries. He consults on statistical modelling in both industry, academia, and non-profits, applying causal inference techniques across scientific disciplines. He currently works in the pharmaceutical sector.Personal website: www.theissbendixen.com Benjamin Grant Purzycki is Associate Professor at Aarhus University. He is a cognitive and evolutionary cultural anthropologist and focuses on the causal role of various demographic and cultural factors on human cooperation. He has conducted fieldwork in the Tyva Republic (Russia) and managed large, cross-cultural projects. His most recent books include The Minds of Gods: New Horizons in the Naturalistic Study of Religion (Bloomsbury), Ethnographic Free-List Data (Sage), and Morality and the Gods (Cambridge University Press).Personal website: www.bgpurzycki.wordpress.com

The Data Analyst’s Guide to Cause and Effect offers an excellent, comprehensive, yet accessible introduction to causal inference. With a light-hearted approach, it opens up a new perspective for those accustomed to traditional statistical analysis, shedding light on crucial aspects of data interpretation. From selecting the right controls to estimating causal effects and even tackling advanced topics like missing data and the intricacies of multilevel modeling, this book is an invaluable guide for analysts seeking to move beyond mere correlation.

• About the AuthorsSeries Editor’s IntroductionAcknowledgmentsChapter 1: IntroductionThe fundamental promise of causal inferenceCausal inference is “EEESI”The R programming languageFormal notationChapter objectivesFurther readingChapter 2: Causal GraphsRandomizing a DAGElementary ingredients of DAGsGood and bad controlsWhere do DAGs come from?Average people and people on averageChapter objectivesFurther readingChapter 3: G-methods and Marginal EffectsInverse probability weightingG-computationIt’s assumptions all the way downChapter objectivesFurther readingChapter 4: Adventures in G-methodsDoubly robust estimationSub-group analysisComplex longitudinal designsMediation analysis: Crossing hypothetical worldsChapter objectivesFurther readingChapter 5: Most of Your Data is Almost Always MissingExternal validity and selection biasPoststratificationThe treatment effects zooTarget populations and econometricsChapter objectivesFurther readingChapter 6: More Missing DataTo be or not to be missingCompletely random terminologyMissing data imputationChapter objectivesFurther readingChapter 7: Multilevel modelling and Mundlak’s legacyCausal inference as counterfactual predictionMundlak modelsMarginal effects in a multilevel modelChapter objectivesFurther readingChapter 8: Causal Inference is not EasyViolations of identification assumptions and some solutionsBayesian causal modellingPerspectives on RCT data analysisCausal inference in the era of Big Data and AIConclusionReferencesIndex