Back to Lessons Confidence Intervals 0 pts Module 3 · Lesson 2
Introduction

Confidence Intervals

What they tell you that p-values don't

The Problem with P-Values

A p-value tells you one thing: is this result likely due to chance? It says nothing about how big the effect is, how precise your estimate is, or whether the effect matters clinically. Confidence intervals fix this.

What You'll Learn

How to read confidence intervals to understand the range of plausible effects, why width matters as much as the point estimate, and how to use CIs to make better clinical decisions than p-values alone would allow.

What a Confidence Interval Actually Means

1

The Definition

A 95% confidence interval means: if we repeated this study 100 times, about 95 of those intervals would contain the true population value.

It's a range of plausible values for the true effect, given your data.

2

What It Contains

Point estimate: Your best guess (the middle of the CI)
Lower bound: The smallest plausible effect
Upper bound: The largest plausible effect

3

The Key Insight

The true effect could be anywhere in that range. A CI of 2% to 40% is very different from 18% to 22%, even if both have the same point estimate of 20%.

The confidence interval tells you what you don't know. A wide CI means high uncertainty; a narrow CI means you've pinned it down.

Width Reflects Precision

The width of a CI tells you how precise your estimate is—which depends mainly on sample size.

Same point estimate (20% reduction), different sample sizes:

N = 50
-10% to 50%
N = 200
5% to 35%
N = 1,000
13% to 27%
Red line = null (no effect). Notice how only the large study excludes zero entirely. The small study is compatible with both harm and benefit.

A wide CI doesn't mean no effect—it means we don't know yet. A narrow CI around zero means we're confident there's no meaningful effect.

CIs Tell You More Than P-Values

1

Significance from CI

If the 95% CI excludes the null value (0 for differences, 1 for ratios), the result is significant at p < 0.05. You get significance information for free.

2

Effect Size from CI

The point estimate tells you the magnitude. P-values don't.

3

Precision from CI

The width tells you how confident to be. P-values don't distinguish between precise and imprecise estimates.

4

Clinical Relevance from CI

You can see if the entire CI is above or below clinically meaningful thresholds. P-values can't do this.

Every p-value can be converted to a CI, but not vice versa. The CI is strictly more informative.

Four Patterns to Recognize

1

Significant and Precise

Effect: 25%, 95% CI: 20% to 30%

CI excludes zero, narrow width. Strong evidence of a real, clinically meaningful effect.

2

Significant but Imprecise

Effect: 25%, 95% CI: 2% to 48%

CI excludes zero (significant), but the true effect could be anywhere from trivial (2%) to huge (48%). Needs more data.

3

Not Significant, Inconclusive

Effect: 15%, 95% CI: -10% to 40%

CI includes zero, but also includes clinically meaningful effects. We can't rule out benefit OR harm. Underpowered.

4

Not Significant, Definitive

Effect: 0.5%, 95% CI: -2% to 3%

CI includes zero but is narrow. Even the extremes are clinically irrelevant. This is a confident null—no meaningful difference.

The Overlapping CI Trap

A common mistake: "The CIs overlap, so the groups aren't different."

!

This Is Often Wrong

Two groups can have overlapping CIs and still be significantly different from each other. The correct comparison is the CI of the difference between groups, not whether individual CIs overlap.

Example

Group A: 40% success, 95% CI: 30% to 50%
Group B: 55% success, 95% CI: 45% to 65%

The CIs overlap (both include 45-50%), but the difference (15%) has its own CI: 2% to 28%. Since this excludes zero, the groups are significantly different.

Never judge significance by eyeballing overlap. Look for the CI of the difference, or the p-value for the comparison.

Exercise: Interpret the CI

For each scenario, choose the best interpretation.

Question 1 of 8

The Bottom Line

0
Total Points Earned
Exercise (0/8 correct) +0 pts
Lesson Completed +100 pts
  • CIs give you the range of plausible effects. The true value could be anywhere in that range.
  • Width reflects precision. Narrow = confident estimate. Wide = uncertain, need more data.
  • CIs tell you more than p-values. Effect size, precision, and significance—all in one measure.
  • A narrow CI around zero is informative. It's a confident null—you've proven equivalence.
  • Don't judge by overlapping CIs. The CI of the difference is what matters for comparisons.

Your New Habit

When reading a study, look at the CI before the p-value. Ask: "What's the smallest plausible effect? What's the largest? Is this range narrow enough to be informative?"