Course contentsShow
← Data Science
Lesson 1439 of 2,14536. Causal Inference BasicsFree lesson

Intent-to-Treat Analysis

Analyzing all participants as randomized regardless of compliance, and why this preserves causal interpretation.

Intent-to-Treat Analysis

You'll learn why analyzing participants exactly as randomized—even when they don't follow the protocol—protects causal conclusions in experiments.

The Core Principle

In a Randomized Controlled Trial (RCT), you carefully assign participants to treatment or control groups through randomization. But what happens when participants don't comply? Some assigned to treatment never take it. Others in control somehow get the treatment. Some drop out entirely.

Intent-to-Treat (ITT) means you analyze every participant in the group they were originally randomized to, regardless of what they actually did. If someone was randomized to receive a new drug but never took a single pill, you still count them in the treatment group for analysis.

Why This Matters for Causality

Randomization is your guarantee that treatment and control groups are comparable at baseline—any differences you observe can be attributed to the treatment itself, not to pre-existing differences. But the moment you start excluding non-compliers or reclassifying people based on what they actually did, you break randomization.

Analogy: Imagine flipping a fair coin to assign students to tutoring. The coin ensures both groups are equivalent. But if you later move students who skipped tutoring sessions into the control group, you're no longer comparing "tutoring vs. no tutoring." You're comparing "motivated students who attended vs. everyone else"—and motivation is a confounder you've just introduced.

ITT preserves the original randomization, maintaining the causal interpretation. Yes, it may dilute the estimated effect (since some in the treatment group didn't actually receive treatment), but it gives you an honest answer to: "What is the effect of being assigned to this intervention?"

Key Takeaway: Intent-to-Treat analysis maintains the causal validity of RCTs by analyzing participants according to their randomized assignment, preventing selection bias that would occur if we excluded or reclassified based on compliance.