A 90% confidence interval for a proportion is (0.42, 0.58). What is its margin of error? Give the exact decimal.
Confidence intervals
Estimate ± multiplier × standard error for and : the three spreads, and , margin of error, interpreting an interval in context, and when to use an interval instead of a test.
Key ideas
- Second limitation of tests
A large p-value does not show is true. If the goal is a plausible value for the parameter, use an interval.
- General form
statistic ± multiplier × standard error. The margin of error is multiplier × SE, half the interval's width, always positive.
- Standard error
For a CI, the parameter is unknown, so the spread is estimated from the sample: or .
- Multipliers
Proportions use , the same for every : 1.645 (90%), 1.96 (95%), 2.576 (99%). Means use with , which changes with even at the same level.
- Interpretation
"We are [level]% confident that [the parameter, in context, with population] is between [a] and [b] [units]." The interval estimates a parameter, not individual values.
- Test or interval?
A yes/no research question (is it different from, more than…) calls for a test. A "what is the value / how large" question calls for an interval.
Formulas (on the Exam 2 help card)
The three spreads
| Spread | Formula for proportions | Used for |
|---|---|---|
| Standard deviation | describing when is known | |
| Null standard deviation | hypothesis tests (assume ) | |
| Standard error | confidence intervals (estimate ) |
Two intervals
(1) A random sample of 800 county residents finds 520 support a new transit levy. Build a 95% interval for the support rate. (2) A random sample of 30 nurses reports sleep before a shift. Interpret the interval in the output.
t.test(sleep_hours, conf.level = 0.95)
One Sample t-test
data: sleep_hours
t = 24.83, df = 29, p-value = 4.338e-21
alternative hypothesis: true mean is not equal to 0
95 percent confidence interval:
6.239891 7.360109
sample estimates:
mean of x
6.8 (1) Statistic: .
SE: . Margin: .
Interval: . We are 95% confident that the true proportion of all county residents who support the levy is between 0.617 and 0.683.
(2) The center is and the multiplier came from a t-distribution with 29 df. We are 95% confident that the true mean pre-shift sleep of all nurses in this population is between 6.24 and 7.36 hours.
Ignore the test line: without
mu =, R tests , which no one asked about. Only the interval answers this question.
Support: (0.617, 0.683). Sleep: (6.24, 7.36) hours. Both describe a population parameter, not individual residents or nurses.
Check yourself
A confidence interval for a mean commute is (12.4, 15.0) minutes. What sample mean was it built around? Give the value to one decimal place.
Which spread belongs in a confidence interval for a population proportion?
A 99% confidence interval for a proportion uses which multiplier?
A city asks: "What is the average number of hours per week our residents volunteer?" Which procedure answers this?
A lab has a 95% confidence interval for the mean weight of its packages. Select every change that would make its next interval narrower, assuming nothing else changes.
An interval for a mean uses and a standard error of 1.5 grams. What is the margin of error, in grams? Give the exact value.