Lesson 4 of 7 · 15 min

Margin of error and reading an interval

Calculate the margin of error, read p̂ and the margin straight from a given interval, and work backwards to the sample size or confidence level behind it.

What you will be able to do

  • Calculate the approximate margin of error E=zp^(1−p^)nE = z\sqrt{\frac{\hat{p}(1-\hat{p})}{n}}.
  • Find p^\hat{p} and EE from the endpoints of an interval.
  • Work backwards from an interval to the sample size or the confidence level.

When a news report says a poll has "a margin of error of 3 percentage points", it is describing half the width of a 95% confidence interval. The margin of error is the "±" part, and it is the single number that tells you how precise an estimate is.

Definition

Margin of error

E=zp^(1−p^)nE = z\sqrt{\frac{\hat{p}(1-\hat{p})}{n}}

The confidence interval is (p^−E, p^+E)(\hat{p} - E,\ \hat{p} + E), so it has width 2E2E and is centred on p^\hat{p}.

The anatomy of a confidence interval: centre p^\hat{p}, half-width EE on each side, total width 2E2E.

Key idea

Reading an interval (a, b)

p^=a+b2E=b−a2\hat{p} = \frac{a + b}{2} \qquad E = \frac{b - a}{2}

The centre is the sample proportion and half the width is the margin of error.

A confidence interval for pp is (0.552,0.608)(0.552, 0.608). Find the sample proportion p^\hat{p} and the margin of error EE.

Without a calculator: a random sample of 300 gives p^=0.25\hat{p} = 0.25. Find the margin of error for a 95% confidence interval, using z=1.96z = 1.96.

The margin of error links four quantities: EE, zz (the confidence level), p^\hat{p} and nn. Know any three and you can find the fourth. Two "backwards" questions come up often.

Worked example

Work back from an interval to the sample size

  • Complex familiar
  • Technology active

A survey of Australian university students reports a 95% confidence interval of (0.412,0.468)(0.412, 0.468) for the proportion who work more than 15 hours a week. Find the sample proportion, the margin of error and the approximate number of students surveyed.

5 steps

Try it yourself first, then check the working one step at a time.

Method

  1. p^\hat{p} = midpoint, EE = half the width.

  2. n=(zE)2p^(1−p^)n = \left(\frac{z}{E}\right)^2\hat{p}(1-\hat{p}).

  3. Round to a sensible value: the endpoints were rounded.

Watch out for

  • Using the full width 0.056 as EE, which gives a quarter of the sample size.

Worked example

Find the confidence level of an interval

  • Complex familiar
  • Technology active

In a random sample of 600 Gold Coast residents, 210 said they surf at least once a month. A report gives the interval (0.3118,0.3882)(0.3118, 0.3882) for the proportion of all Gold Coast residents who surf at least monthly, but forgets to say the confidence level. Determine the confidence level.

4 steps

Try it yourself first, then check the working one step at a time.

Method

  1. EE = half the width; p^\hat{p} from the data.

  2. z=E÷p^(1−p^)nz = E \div \sqrt{\frac{\hat{p}(1-\hat{p})}{n}}.

  3. C=P(−z<Z<z)C = P(-z < Z < z), from normal cdf.

Watch out for

  • Finding P(Z<z)=0.975P(Z < z) = 0.975 and reporting 97.5%: the confidence level is the central area.

Common mistake

Endpoints are rounded, so backwards answers are approximate

Interval endpoints are usually rounded to 3 decimal places, so EE read from them is slightly off, and so is any nn or CC you calculate. If you get n=1207.4n = 1207.4, the sample was about 1200; if you get z=1.962z = 1.962, the level was 95%. Round to a sensible value and say so.

From the same sample, three intervals for pp are calculated at 90%, 95% and 99% confidence: (0.42,0.50)(0.42, 0.50), (0.427,0.493)(0.427, 0.493) and (0.408,0.512)(0.408, 0.512). Which one is the 99% interval?

Choose one answer

Summary

  • The margin of error is E=zp^(1−p^)nE = z\sqrt{\frac{\hat{p}(1-\hat{p})}{n}}, half the width of the interval.

  • For an interval (a,b)(a, b): p^=a+b2\hat{p} = \frac{a+b}{2} and E=b−a2E = \frac{b-a}{2}.

  • To find nn: solve E=zp^(1−p^)nE = z\sqrt{\frac{\hat{p}(1-\hat{p})}{n}}, giving n=(zE)2p^(1−p^)n = \left(\frac{z}{E}\right)^2\hat{p}(1-\hat{p}).

  • To find the level: z=Ep^(1−p^)/nz = \frac{E}{\sqrt{\hat{p}(1-\hat{p})/n}}, then C=P(−z<Z<z)C = P(-z < Z < z).

  • Backwards answers inherit rounding error from the endpoints, so round them sensibly.

Finished the lesson?

Try the 3 quick checks above and the lesson completes itself, or mark it complete now.

Lesson 4 of 70% read

Save your progress with a free account