Course contentsShow
← Data Science
Lesson 1440 of 2,14536. Causal Inference BasicsFree lesson

Treatment Effect Estimation

Calculating average treatment effects (ATE) by comparing outcomes between treatment and control groups.

Treatment Effect Estimation

You'll learn how to quantify the actual impact of a treatment by comparing outcomes between treatment and control groups.

The Core Idea

After running a randomized controlled trial (which you just learned creates balanced groups through randomization), you need to measure how much the treatment actually changed the outcome. Treatment Effect Estimation calculates the difference in average outcomes between those who received the treatment and those who didn't.

The most common metric is the Average Treatment Effect (ATE), which answers: "On average, how much did the treatment change the outcome compared to no treatment?"

Calculating ATE

The formula is beautifully simple:

ATE = (Average outcome in treatment group) - (Average outcome in control group)

For example, imagine testing a new drug to lower blood pressure:

  • Treatment group (received drug): average BP reduction = 15 mmHg
  • Control group (received placebo): average BP reduction = 3 mmHg
  • ATE = 15 - 3 = 12 mmHg

The drug caused an additional 12 mmHg reduction on average.

Why This Works in RCTs

Because randomization created balanced groups (no confounders), any systematic difference in outcomes can be attributed to the treatment itself. You're comparing "apples to apples" — groups that differ only in whether they received the treatment.

Beyond Simple Averages

While ATE gives you the population-level average effect, individual responses may vary. Some people might benefit more, others less. But ATE provides the single most important summary statistic for policy and decision-making: the expected effect if you applied this treatment broadly.

Key Takeaway: The Average Treatment Effect (ATE) quantifies causal impact by subtracting the control group's average outcome from the treatment group's average outcome — a simple but powerful measure enabled by proper randomization.