Four planned one-proportion z-tests. Select every plan whose sample is large enough for the z-test.
The z-test for a proportion
Theory-based tests for : hypotheses, the null standard deviation, the success-failure check with , interpreting z and the p-value, and three-part conclusions.
Key ideas
- Null standard deviation
: how much would wiggle if were true. A test assumes , so it uses , not .
- Sample size condition
and , using the null value. Using can wrongly pass or fail the check.
- Test statistic
"The observed sample proportion of [context] is |z| null standard deviations above/below the hypothesized value ." The sign comes from , not from .
- p-value definition (three ideas)
(1) Assuming is true, in context; (2) a statistic as extreme as or more extreme than the one observed, in the direction of ; (3) its probability.
- Conclusion (three elements)
(1) The strength of evidence; (2) the parameter and its population; (3) the direction of . Always phrased about , even when the p-value is large.
- Two-sided p-values
Shade both tails: double the one-tail area beyond . Switching to a one-sided halves it only if the new direction matches the data.
Formulas
Three spreads that look alike
| Name | Formula | Uses | When |
|---|---|---|---|
| Standard deviation | true | describing sampling variability when is known | |
| Null standard deviation | null value | hypothesis tests (z) | |
| Standard error | sample | confidence intervals (Exam 2) |
Phone payments at one station
A transit agency reports that 45% of riders pay by phone. A station manager suspects the rate is higher at her station. In a random sample of 400 riders there, 204 paid by phone. Test her claim.
> phat <- 204 / 400 > z <- (phat - 0.45) / sqrt(0.45 * 0.55 / 400) > z [1] 2.412091 > pnorm(z, 0, 1, lower.tail = FALSE) [1] 0.007930666
Parameter and hypotheses: = the true proportion of riders at this station who pay by phone. versus .
Condition with $\pi_0$: and , both at least 10.
Statistic: ; null SD ; : the sample proportion is about 2.41 null SDs above 0.45.
p-value: right tail, about 0.0079. Assuming 45% of riders at this station pay by phone, there is about a 0.79% chance of a sample proportion of 0.51 or higher in a sample of 400.
Conclusion: .
Very strong evidence that the true proportion of riders at this station who pay by phone is greater than 0.45.
Check yourself
For versus about the share of a clinic's patients who arrive on time, the test gives . A student writes: "The observed proportion of on-time patients is 1.9 standard errors below 0.70." What, if anything, needs fixing?
A two-sided z-test gives . In R, pnorm(-2.1, 0, 1) returns 0.0179. What is the p-value?
Testing with a random sample of 100 gives . Compute the z-statistic. Give the exact value (or 2 decimal places).
A z-test of versus about city residents who bike to work gives p = 0.07. Select every element a complete STATS 250 conclusion needs.
Quantity A: the p-value of a right-tailed z-test with , and . Quantity B: the p-value with the same and but .
A test of versus about the defect rate of a factory's phone screens gives and p = 0.04. Which conclusion is correct?