Category
Experimental Methods & Causal Inference
37 free guides in this category.
13 guides
Causal Diagrams (DAGs)
Collider Bias & Selection Bias
Discover how conditioning on a common outcome can create false correlations and distort your view of reality.
Read guideCommon Causes (Confounders)
Understand how a hidden third factor can create a misleading link between two unrelated things.
Read guideCommon DAG Mistakes
Avoidable pitfalls when drawing and interpreting causal diagrams that can lead to biased conclusions.
Read guideCorrelation vs. Causation
Just because two things happen together doesn't mean one causes the other—this critical distinction helps you avoid being misled by coincidences in data.
Read guideDAGs and Regression Models
Use causal diagrams (DAGs) as a blueprint to build more accurate and less biased statistical regression models.
Read guideFeedback Loops and Cycles
Learn to identify and map the chains of cause-and-effect that amplify or stabilize systems, leading to better predictions and interventions.
Read guideM-Bias
A subtle form of bias that occurs when you mistakenly control for a variable that appears to be a confounder but is actually a collider, opening spurious causal pathways.
Read guideMeasurement Error
When the data we measure differs from reality, measurement error in our causal diagrams can hide, create, or reverse relationships between variables.
Read guidePrediction vs. Causation in DAGs
DAGs can predict outcomes from patterns, but they cannot prove that one thing causes another without additional assumptions about how variables truly relate.
Read guideStructural Causal Models (SCMs)
A powerful framework that combines diagrams and simple rules to represent causal relationships and predict the effects of interventions.
Read guideTesting Causal Assumptions
Methods for verifying whether causal model assumptions hold in observational data, including conditional independence tests and falsification checks.
Read guideCausal Sufficiency Assumption
The critical assumption that all common causes influencing both a treatment and an outcome have been measured and included in your analysis.
Read guidePositivity Assumption
The requirement that every individual must have had some chance of receiving each treatment option for valid causal comparison.
Read guide6 guides
Causality & Counterfactuals
Counterfactuals and DAGs
Using visual maps of causes (DAGs) to rigorously answer 'what if' questions about events that never happened.
Read guideDAGs vs. Potential Outcomes
Two complementary frameworks for understanding cause and effect: one maps relationships (DAGs), the other imagines 'what if' scenarios (Potential Outcomes).
Read guideExchangeability & Ignorability
Understand the 'apples-to-apples' principle crucial for determining true cause and effect in any comparison.
Read guidePotential Outcomes Framework
A framework for defining causal effects by comparing what actually happened to what would have happened under different circumstances.
Read guideSingle World Intervention Graphs
A graph that shows counterfactual variables for a single intervention choice, making “what if” reasoning precise.
Read guideCausal Effect
The specific impact an action has on an outcome, measured by comparing what happened to what would have happened without that action.
Read guide3 guides
Confounding & Bias
Internal vs External Validity
Internal validity asks if a study's design supports causal conclusions, while external validity asks if those findings apply beyond the study to real-world settings.
Read guideTime-Varying Confounding
A special type of confounding where variables that influence both exposure and outcome are themselves affected by earlier exposure, creating a dilemma for standard adjustment methods.
Read guideUnobserved Confounding
The error of assuming a direct link between two things when a hidden, unmeasured factor is actually influencing both.
Read guide4 guides
Hypothesis Testing & Statistical Power
Heterogeneous Treatment Effects
The study of how treatments affect different people or groups in different ways, moving beyond simple averages to understand who benefits most.
Read guidePower Analysis & Sample Size
A statistical method for calculating the minimum sample size required for a study to have a high probability of detecting a true effect of a specified size.
Read guidePre-registration & Analysis Plans
Documenting your hypotheses and analysis methods before collecting data helps you distinguish genuine discoveries from lucky patterns found after the fact.
Read guideSequential Testing
A flexible approach to decision-making that lets you analyze evidence as it arrives and stop early when the answer becomes clear.
Read guide3 guides
Quasi-Experimental Designs
Difference-in-Differences
A method for estimating causal effects by comparing changes over time between a treatment group and a control group.
Read guideInstrumental Variables
A clever method for uncovering true cause-and-effect relationships when you can't run a controlled experiment.
Read guideRegression Discontinuity Design
A research method that estimates causal effects by comparing outcomes just above and below a treatment threshold.
Read guide5 guides
Randomized Controlled Trials (RCTs)
Blocking & Stratified Randomization
Advanced randomization techniques that balance groups across important factors and prevent chance imbalances during participant assignment.
Read guideDesigning Outcomes and Metrics
Learn how to define clear, measurable goals for any experiment or project to know if you're truly succeeding.
Read guideResearch Ethics & IRB
The essential framework of ethical guidelines, informed consent procedures, and Institutional Review Board oversight that protects human participants in research studies.
Read guideRandomized Experiments
A method for determining cause-and-effect by randomly assigning subjects to treatment or control groups.
Read guideUnit of Randomization & Clustering
Understand why choosing what to randomize—individuals or groups—is a critical decision that can make or break an experiment.
Read guide3 guides
Threats to Validity
Attrition & Missing Data
When participants drop out of studies or data goes missing, the remaining group may no longer represent the original sample, potentially distorting conclusions about cause and effect.
Read guideInterference, Spillovers, SUTVA
When one person's treatment affects another's outcome, challenging core assumptions about causal effects in experiments.
Read guideNoncompliance and Intent-to-Treat
When participants switch treatments or drop out, special analysis methods like Intent-to-Treat and Instrumental Variables help recover the true causal effect.
Read guide