Name: _________________________________ Period: _____________ Date: _________________

Testing a Claim Practice

1. Hip replacements require surgery followed by a hospital stay to monitor for complications. A doctor has devised a
new surgical procedure that she believes will reduce the mean length of hospital stay from its present value of 3 days.
Although the procedure itself is more expensive, if successful, the procedure will reduce the overall cost because of a
shorter hospital stay. Define the parameter of interest, state the hypotheses, describe Type I and Type II errors and discuss
the consequences of each type of error in terms of the cost.
Let µ = the mean length of a hospital stay (in days) after hip replacements using the new procedure. H0: µ = 3 vs. Ha: µ < 3.
Type I error: Finding convincing evidence that the mean hospital stay has decreased when it really hasn’t. A consequence of
this type of error is that the hospital will use this more expensive procedure even though it is not effective. Also, the hospital
might not investigate other ways to reduce the mean hospital stay. Some patients may be released too early.
Type II error: Not finding convincing evidence that the mean hospital stay has decreased when it really has. A potential
consequence is that the hospital will not use the new procedure, even though it is effective. Also, the patients might stay in
the hospital longer than necessary, increasing their hospital costs.
The hospital does not want to release patients too early. Would they prefer a significance level of .1 or .01? Briefly explain.
If it is more important to avoid a Type I error, choose a small value of α. If it is more important to avoid a Type II error, choose
a high value of α. To reduce the probability of rejecting H0 and releasing patients too early, the hospital should choose 0.01.

2. LeJon Brames, a starting player on a local basketball team, made only 41% of his free throws last season. Hoping to
improve his chances at landing a college scholarship, he worked on developing a softer shot to improve his free throw
accuracy. In the following season, LeJon made 18 free throws in a random sample of 30 attempts.
a. Does the sample provide convincing evidence that LeJon has improved? Use α = .02.
Identify: Let ρ = the true proportion of free throws attempts that LeJon makes during the season following his training.
We want to test the following at the 𝛼 = 0.02 significance level: H0: ρ = 0.41 Ha: ρ > 0.41
Conditions: We will perform a one-sample z test for proportion, if the following conditions are met:
✓ Random: 30 free throw attempts were randomly selected.
✓ 10%: It is reasonable to assume that 30 is less than 10% of all of his free throw attempts the following season.
✓ Large counts: Assuming H0 is true, p0 = 0.41. Thus, np0 = 30(0.41) = 12.3 ≥ 10 and n(1-p0) = 30(0.59) = 17.7 ≥ 10.

Calculate: Because the hypothesized value is p0 = 0.41 and p̂ = 18/30 = 0.60, the standardized test statistic is
pˆ − p 0 0.60 − 0.41 (Note: You must use hypothesized value p0 = 0.41 in calculation of standard deviation).
z= = = 2.116
p 0(1 − p 0) 0.41(1 − 0.41)
n 30

Using Table A, P(Z ≥ 2.12) = 1 – P(Z < 2.12) = 0.017 where Z has a standard normal distribution.
Using technology, Normalcdf(lower: 2.116, upper: 1000, µ = 0, σ = 1) gives a P-value of 0.0172. So, P(Z ≥ 2.116) = 0.0172.
Interpret: Because the P-value (0.0172) is less than α (0.02), we reject H0. There is convincing (or statistically significant)
evidence that the proportion of free throw attempts that LeJon makes is more than 0.41.
b. What is the probability of a Type I error? How do you interpret the P-value? P(Type I error) = α = 0.02. The P-value is the
probability, assuming that H0 is true, that the sample statistic (e.g., p̂ ) would be as extreme or more extreme than the actually
observed outcome. Here, assuming that the mean proportion of successful free throw attempts is p = 0.41, there is a 0.0172
probability of getting a sample proportion of p̂ = 18/30 = 0.60 or higher by chance alone in a random sample of 30 attempts.
c. What type of error might you have made? Explain the mistake in context.
Since we rejected H0, we might have made a Type I error. That is, we believe ρ > 0.41 when in fact ρ = 0.41.
d. Suppose you want to know whether LeJon’s success rate changed (a 2-tailed test with α = 0.02). State the new Ha and
explain how the P-value and conclusion would change? Ha would become ρ ≠ 0.41. Also, we need to double the P-value
from (b) because the area on the table (or from a calculator) is for one-tail, but this is a two-tailed test. P-value: 2P(Z ≥ 2.116)
= 2(0.0172) = 0.0344 where Z has a standard normal distribution. Because the P-value (0.0344) is greater than α (0.02) in the
two-sided test, we fail to reject H0. At this significance level, there is not convincing evidence that LeJon’s success rate is
different from 0.41.
3. Viewers of the TV series Breaking Bad had an average income level of $40,000. To attract a younger audience, the
producers created a sequel: Breaking Bad in Legoland. The producers claim that viewers of the new series have an average
income level greater than $40,000. A random sample of 40 viewers of the new series produces a sample mean income of
$42,000 and standard deviation of $9,000.
a. Does the evidence support the producers’ claim? Use a significance test with α = 0.05 to find out.
Identify: Let µ = the true mean income of viewers of the sequel. We want to test the following at α = 0.05:
H0: µ = 40,000 Ha: µ > 40,000
Conditions: We will conduct a one-sample t test for mean if the following conditions are met:
✓ Random: A random sample of 40 viewers was selected.
✓ Independence/10%: It is reasonable to assume that there are more than 10∙(40) = 400 viewers of the sequel. After all,
any movie about Legoland is sure to be a … blockbuster.
✓ Normal/Large sample: The sample size is large (n = 40 ≥ 30). The large sample size makes it OK to use t procedures.

x −  42,000 − 40,000
Calculate: t = = = 1.405 (Note: With t-statistics, you must state the degrees of freedom.)
s 9000
n 40
• The calculator’s 1 sample t-test for mean with df = 39 reveals P-value = 0.0839.
• tcdf(lower: 1.405; upper: 1000, df: 39) gives a P-value of 0.0839. So, P(t > 1.405) = 0.0839.
Interpret: Since the P-value (0.0839) is greater than α (0.05), we fail to reject the null hypothesis. There is not
convincing (or statistically significant) evidence that mean income of viewers of the sequel is greater than $40,000.
b. What conclusion would you make if α = 0.10?
Since the P-value (0.0839) is now less than α (0.10), we reject the null hypothesis at this significance level. There is convincing
(or statistically significant) evidence that mean income of viewers is greater than $40,000.
c. Describe a Type I and Type II error in this situation.
Type I: Finding convincing evidence that µ > $40,000, when in fact µ = $40,000.
Type II: Failing to find convincing evidence that µ > $40,000, when in fact µ > $40,000.
d. Suppose the true mean income is µ = 44,000. When α = 0.05, the power of the test is 0.878. Interpret this value.

Given that µ = 44,000, there is a 0.878 probability of finding convincing evidence that µ > 40,000.
e. What is β (the probability of a Type II error) for this test? Interpret this value in context.

P(Type II) = 1- Power = 1 – 0.878 = 0.122 = β. There is a 0.122 probability of not finding convincing evidence to believe that µ >
40,000, when in reality µ > 40,000.

β
Label/shade α, β, and power on the sketch to the right.
Power
f. Identify three (or more) ways to increase the power of the test.
Explain how each increases power.
1. Increase the sample size. This reduces the standard deviation of the sampling distribution. With a smaller standard error, a
test is more likely to be able to detect a difference in results from the two treatments, if such a difference exists.
2. Increase significance level, α. If α is larger, it is easier to reject H0, which increases the power of the test. However,
increasing α increases the probability of a Type I error.
3. Increase the distance between H0 and Ha values (i.e., increase the “effect size”), if possible. This makes it easier to detect a
difference.
4. Decrease the standard error by using a different sampling method or other design improvements.
 Name: _________________________________ Period: _____________ Date: _________________
Testing a Claim Practice
4. A random sample of students from a large high school were chosen to determine if their sitting pulse rate was
different from their standing pulse rate. Each student’s pulse rate was measured in both positions. The order of the
measurement positions was determined randomly.

Sitting 62 74 82 88 82 66 64 84 72 82 80 72 64 62
Standing 68 78 80 92 58 96 72 100 82 76 92 74 60 58
Difference -6 -4 2 -4 24 -30 -8 -16 -10 6 -12 -2 4 4
Can we conclude that there is a difference in pulse rates? Use a level of significance of α = .10.
Identify: Let µdiff = the true mean difference (Sitting - Standing) in pulse rates. In this case, a positive difference means that
the sitting pulse rate is higher (and the standing pulse rate is lower).
We want to test the following at the α = 0.10 significance level: H0: µdiff = 0 Ha: µdiff ≠ 0.
Conditions: If conditions are met, we will perform a paired t test for the difference in means (µdiff).
✓ Random: A random sample of high school students was chosen, and the order of measurement was random.
✓ 10%: It is reasonable to assume that there are more than 14(10) = 140 students in this large high school.
✓ Normal/Large sample: The graphs below do not show much skewness, and there are no outliers. So, it is reasonable
to use t procedures for these data.

Calculate: x diff = -3.7143 sdiff = 12.375

x −  −3.7143 − 0
t= = = −1.123
s 12.375
n 14
Technology: The calculator provides only the areas for a one-tail test. tcdf(lower: -∞, upper: -1.123, df=13) yields P-value of
0.14087477. Because t distributions are symmetric, P(t < - 1.123) = P(t > 1.123). Because this is a two-tailed test, we double
the P-value from the one-tail test to get the P-value for a two-tail test. 2P(t < -1.123) = 2(0.14087477) = 0.2817.
OR
We entered the differences in the calculator and used the calculator’s t-test command. P-value = 0.2818 with df = 14 - 1 = 13.
Interpret: Since the P-value (0.2818) is greater than α (0.10), we fail to reject the null hypothesis. There is not convincing
evidence that the true mean difference in pulse rates is different than zero.
Enter time in express lane in List1 and time in regular lane in List 2. To compute the paired differences, navigate to the heading
of List3 and enter shift List 2 – shift List 1. Using the computed differences, follow the instructions for a 1-sample t-test. Input
Data: List
µ: ≠µ0
µ0: 0
List: List3
Freq: 1 Press EXE, which returns the t-statistic and the probability for a two-tailed test.
Exam tips for students:
• When writing hypotheses statements, defining the parameter or writing your conclusion; do not refer to the sample
statistic (e.g., x , p̂ ), the sample, or the data.
• In your interpretation of a significance test, you should say either that you reject the null hypothesis or that you fail to
reject the null hypothesis. Do not say that a hypothesis is “accepted” or “proven true.”
• If a significance test leads to the decision to fail to reject H0,
o be sure to interpret the result as “we do not have convincing evidence to believe Ha (in context).
o do not say that there is convincing evidence to support the null hypothesis.
o do not say that the true mean or proportion is equal to the hypothesized value.