Course contentsShow
← Data Science
Lesson 271 of 2,1458. Confidence IntervalsFree lesson

Margin of Error

Breaking down the margin of error: the half-width of a confidence interval and factors that influence it.

Margin of Error

What you'll learn: You'll understand what the margin of error is and what makes it larger or smaller in confidence intervals.


What Is the Margin of Error?

The margin of error is the "plus-or-minus" part of a confidence interval—it tells you how far your estimate might reasonably be from the true population parameter. If you see a poll result like "52% ± 3%," that ±3% is the margin of error.

Mathematically, it's exactly half the width of your confidence interval. When you construct a confidence interval for a population mean, you calculate:

Confidence Interval = Sample Mean ± Margin of Error

The margin of error is determined by multiplying the standard error by the critical value (from the t-distribution):

Margin of Error = Critical Value × Standard Error

What Influences the Margin of Error?

Three main factors control how large your margin of error becomes:

  1. Sample size (n): Larger samples → smaller standard error → smaller margin of error. Doubling your sample size doesn't halve the margin of error, but it does shrink it noticeably.

  2. Variability in the data: More spread (higher standard deviation) → larger standard error → larger margin of error. If your data is all over the place, your estimate is less precise.

  3. Confidence level: Higher confidence (e.g., 99% vs. 95%) → larger critical value → larger margin of error. More certainty means casting a wider net.


Analogy

Think of the margin of error as the "cushion" you need around your best guess. If you're trying to guess someone's height while blindfolded, a larger cushion (margin) means you're more confident you've captured the truth—but you're also being less precise.


Key Takeaway: The margin of error is half the width of a confidence interval; it grows with more variability and higher confidence, and shrinks with larger sample sizes.