Justifying claims from a mean CI: worksheet with answer key

8 problems on justifying claims from a mean CI:, ready to print for class or homework. The answer key prints on its own page, so hand out everything before it.

Problem 1

An orchard weighs a random sample of 30 apples from one harvest and reports a 95% confidence interval for the true mean weight of (152.4, 161.6)(152.4,\ 161.6) grams. A dotplot of the 30 weights is roughly symmetric with no outliers.

(a) Find the sample mean and the margin of error.

Say whether each statement below is justified by this interval, and correct the ones that are not.

  1. About 95% of apples in this harvest weigh between 152.4 g and 161.6 g.
  2. There is a 95% probability that the true mean weight is between 152.4 g and 161.6 g.
  3. We are 95% confident that the interval from 152.4 g to 161.6 g captures the true mean weight of all apples in this harvest.
  4. A true mean weight of 158 g is plausible.
  5. The interval proves that the true mean weight is not 150 g.
  6. There is convincing evidence that the true mean weight of apples in this harvest exceeds 150 g.

Problem 2

A supplement label states that each capsule contains 500 mg of magnesium. A testing lab assays a random sample of 20 capsules from Brand A and finds xˉ=493\bar{x} = 493 mg with s=11s = 11 mg. A dotplot of the 20 assays is roughly symmetric with no outliers.

(a) Construct a 95% confidence interval for the true mean magnesium content of Brand A capsules, and say whether the labeled 500 mg is plausible.

(b) The lab assays 20 capsules of Brand B, which carries the same 500 mg label, and finds xˉ=496\bar{x} = 496 mg with the same s=11s = 11 mg. Build that interval and answer the same question.

(c) A colleague reads part (b) and concludes that the interval shows Brand B capsules really do average 500 mg. Correct him.

Problem 3

An environmental lab wants to know whether two instruments read nitrate concentration differently. Each of 8 randomly selected water samples is run through both instruments. Concentrations, in mg/L:

SampleInstrument AInstrument B
18.28.0
27.57.7
39.18.9
46.87.0
58.78.4
67.98.1
79.49.1
88.08.1

A dotplot of the differences is roughly symmetric with no outliers.

(a) Explain why this is a paired design, and name the parameter.

(b) Construct a 95% confidence interval for that parameter using d=A−Bd = A - B.

(c) The lab director writes: "The interval contains 0, so we have shown the two instruments agree." Rewrite the conclusion correctly and explain what is wrong with hers.

Problem 4

Nine randomly selected club runners each run the same course twice, once in their usual shoe and once in a new model, with the order of the two runs randomized. Times, in seconds:

RunnerUsualNew
1248244
2262259
3231227
4275270
5219216
6254249
7240237
8266262
9227223

A dotplot of the differences is roughly symmetric with no outliers, and so is each column of times.

(a) Construct a 95% confidence interval for the true mean difference μd\mu_d using d=usual−newd = \text{usual} - \text{new}, and justify a claim about the new shoe.

(b) A student ignores the pairing and treats the 18 times as two independent samples. The summary statistics are xˉ1=246.889\bar{x}_1 = 246.889 with s1=19.069s_1 = 19.069 for the usual shoe and xˉ2=243.000\bar{x}_2 = 243.000 with s2=18.722s_2 = 18.722 for the new one. Build that interval with the conservative dfdf and state the conclusion it would produce.

(c) Both intervals are centered at the same number. Explain why they disagree, and say which analysis is valid here.

Problem 5

A software team recruits 36 volunteers and randomly assigns 18 to each of two versions of a scheduling app, then times how long each volunteer takes to complete a fixed set of tasks. Version A gives xˉ1=24.6\bar{x}_1 = 24.6 minutes with s1=5.2s_1 = 5.2; Version B gives xˉ2=20.1\bar{x}_2 = 20.1 minutes with s2=4.8s_2 = 4.8. Dotplots of both groups are roughly symmetric with no outliers.

(a) Explain why this is a two-sample design and not a paired one.

(b) Construct a 95% confidence interval for μ1−μ2\mu_1 - \mu_2 using the conservative degrees of freedom, and justify a claim about the two versions.

(c) Can the team conclude that Version B causes faster task completion? Can it conclude that the general public would be faster on Version B?

Problem 6

A garden center compares tomato seedlings from two seed suppliers. It takes an independent random sample of 12 seedlings from Supplier 1's delivery and an independent random sample of 15 from Supplier 2's, and measures height, in centimeters, after four weeks in the same greenhouse. Supplier 1 gives xˉ1=28.40\bar{x}_1 = 28.40 with s1=1.80s_1 = 1.80; Supplier 2 gives xˉ2=27.95\bar{x}_2 = 27.95 with s2=1.65s_2 = 1.65. Dotplots of both samples are roughly symmetric with no outliers.

(a) Construct a 95% confidence interval for μ1−μ2\mu_1 - \mu_2 using the conservative degrees of freedom.

(b) The buyer says the interval shows the two suppliers' seedlings grow to the same mean height. Say what the interval does justify and what it does not.

(c) A colleague suggests the study should have been run as a paired design instead. Was pairing available here?

Problem 7

A delivery company tests new routing software on 14 randomly selected routes. Each route is driven once under the old software and once under the new, on comparable days, with the order randomized. For the differences d=old time−new timed = \text{old time} - \text{new time}, in minutes, the company reports xˉd=4.2\bar{x}_d = 4.2 and sd=5.6s_d = 5.6. A dotplot of the 14 differences is roughly symmetric with no outliers.

(a) Construct a 95% confidence interval for μd\mu_d.

(b) Write the full-credit statistical justification for a claim about the new software.

(c) The operations manager has not taken a statistics course and has to decide whether to buy licenses. Write two or three sentences for her that are honest about what the study shows and how uncertain it is.

(d) Name three sentences that must not appear in either version, and say what is wrong with each.

Problem 8

A coating manufacturer claims its product reduces mean corrosion depth by at least 0.5 mm. A materials lab cuts each of 12 randomly selected steel panels in half, randomly assigns one half of each panel to be coated, exposes all 24 halves to salt spray for 30 days, and measures corrosion depth. For d=uncoated−coatedd = \text{uncoated} - \text{coated}, in millimeters, the lab finds xˉd=0.34\bar{x}_d = 0.34 with sd=0.21s_d = 0.21. A dotplot of the 12 differences is roughly symmetric with no outliers.

(a) Explain why the paired procedure is the right one.

(b) Construct a 95% confidence interval for μd\mu_d and say whether there is convincing evidence the coating reduces mean corrosion depth at all.

(c) Is the manufacturer's "at least 0.5 mm" claim plausible at the 95% level?

(d) Redo the interval at 99% confidence. Do your answers to (b) and (c) change?

Justifying claims from a mean CI:: answer key

  1. (a) xˉ=157.0\bar{x} = 157.0 g and ME=4.6ME = 4.6 g. Statements 3, 4 and 6 are justified. Statement 1 describes individual apples instead of the mean, statement 2 attaches a probability to a finished interval whose bounds are already fixed, and statement 5 says "proves" when the interval only shows that 150 g is not a plausible value at this confidence level.

  2. (a) SE=2.4597SE = 2.4597 mg, df=19df = 19, t∗=2.093t^* = 2.093, ME=5.1481ME = 5.1481 mg, giving (487.852, 498.148)(487.852,\ 498.148) mg. 500 mg is above the interval, so it is not plausible and there is convincing evidence Brand A averages under its label. (b) (490.852, 501.148)(490.852,\ 501.148) mg contains 500, so the label stays plausible and there is not convincing evidence against it. (c) The interval rules values out, it never establishes one; 500 mg is only one of many plausible values, and 496 mg is the best estimate.

  3. (a) Paired, because both instruments measure the same 8 water samples; the parameter is μd\mu_d, the true mean difference (A minus B). (b) xˉd=0.0375\bar{x}_d = 0.0375, sd=0.232609s_d = 0.232609, SE=0.08224SE = 0.08224, df=7df = 7, t∗=2.365t^* = 2.365, ME=0.194497ME = 0.194497, so the 95% interval is (−0.157, 0.232)(-0.157,\ 0.232) mg/L. (c) The interval contains 0, so a difference of zero stays plausible and there is not convincing evidence the instruments differ. It does not show they agree: values like −0.15-0.15 and 0.230.23 mg/L are equally plausible.

  4. (a) xˉd=3.8889\bar{x}_d = 3.8889 s, sd=0.781736s_d = 0.781736, SE=0.260579SE = 0.260579, df=8df = 8, t∗=2.306t^* = 2.306, so the 95% interval is (3.288, 4.490)(3.288,\ 4.490) seconds. It excludes 0, so there is convincing evidence the new shoe is faster, by about 3.3 to 4.5 seconds. (b) SE=8.9078SE = 8.9078 s and ME=20.541ME = 20.541 s, giving (−16.652, 24.430)(-16.652,\ 24.430), which contains 0 and would find no evidence of a difference. (c) Same center, standard errors 34 times apart: pairing removes the roughly 19-second runner-to-runner variation. The two-sample interval is invalid here because its independence condition fails, since the same nine runners appear in both columns.

  5. (a) Two-sample: each volunteer uses only one version, so no observation has a partner in the other group; the parameter is μ1−μ2\mu_1 - \mu_2. (b) SE=1.668SE = 1.668 min, conservative df=17df = 17, t∗=2.110t^* = 2.110, ME=3.5195ME = 3.5195, so the 95% interval is (0.981, 8.019)(0.981,\ 8.019) minutes. It excludes 0, so there is convincing evidence Version A takes longer, by about 1.0 to 8.0 minutes on average. (c) Cause yes, generalization no: random assignment supports a causal claim within the study, but volunteers are not a random sample, so the result does not extend to the general public.

  6. (a) SE=0.671937SE = 0.671937 cm, conservative df=11df = 11, t∗=2.201t^* = 2.201, ME=1.478934ME = 1.478934, so the 95% interval is (−1.029, 1.929)(-1.029,\ 1.929) cm. (b) It justifies saying there is not convincing evidence of a difference in mean height, since 0 is plausible. It does not justify saying the means are the same: differences anywhere from −1.03-1.03 to 1.931.93 cm are equally plausible. (c) No. Pairing needs a specific partner for each unit, and two separate batches from two suppliers have none; larger samples, not pairing, would narrow this interval.

  7. (a) SE=1.496663SE = 1.496663 min, df=13df = 13, t∗=2.160t^* = 2.160, ME=3.232792ME = 3.232792, so the 95% interval is (0.967, 7.433)(0.967,\ 7.433) minutes. (b) 95% confident the interval from 0.967 to 7.433 minutes captures the true mean reduction per route; 0 is excluded and both bounds are positive, so there is convincing evidence the new software reduces mean driving time, and the randomized drive order attributes it to the software. (c) It saved 4.2 minutes per route on average and the saving looks real, but the data support anything from roughly 1 to 7.4 minutes, so a decision that needs 5 minutes is not settled. (d) Avoid attaching a probability to the finished interval, avoid "proves" and treating 4.2 as the parameter, and avoid describing individual routes, whose spread is sd=5.6s_d = 5.6 rather than SE=1.4967SE = 1.4967.

  8. (a) Paired: the two halves of one panel share the steel and its history, so each coated half has a specific partner; the parameter is μd\mu_d. (b) SE=0.060622SE = 0.060622 mm, df=11df = 11, t∗=2.201t^* = 2.201, ME=0.133429ME = 0.133429, so the 95% interval is (0.207, 0.473)(0.207,\ 0.473) mm. It excludes 0, so there is convincing evidence the coating reduces mean corrosion depth by about 0.21 to 0.47 mm. (c) No. The whole interval sits below 0.5, so "at least 0.5 mm" is not plausible at 95% and there is convincing evidence against it. (d) At 99%, t∗=3.106t^* = 3.106 and the interval is (0.152, 0.528)(0.152,\ 0.528) mm. Part (b) is unchanged, but part (c) reverses: 0.5 now lies inside, so the claim is plausible at 99%.

The key gives final answers only. The full worked solution for each problem is on the matching practice set, and every other sheet is on the worksheets page.

More worksheets from this unit