Statistics Assignment: Confidence Intervals and Hypothesis Testing
Complete solutions with key working and supporting graphs
Q1. Asymmetric confidence interval for μ
Given: P[-zα1 < (x̄ - μ)/(σ/√n) < zα2] = 1 - α, where α1 + α2 = α.
Step 1: Multiply through by the positive quantity σ/√n.
-zα1(σ/√n) < x̄ - μ < zα2(σ/√n)
Step 2: Rearrange the left and right inequalities.
x̄ - zα2(σ/√n) < μ < x̄ + zα1(σ/√n)
Therefore, the 100(1-α)% confidence interval is:
( x̄ - zα2·σ/√n , x̄ + zα1·σ/√n )
Part (b): α = 0.09, α1 = α/3 = 0.03, α2 = 2α/3 = 0.06.
CI = ( x̄ - z0.06·σ/√n , x̄ + z0.03·σ/√n )
z0.06 ≈ 1.555, z0.03 ≈ 1.881
CI = ( x̄ - 1.555·σ/√n , x̄ + 1.881·σ/√n )
Part (c): If the interval is symmetric, then α1 = α2 = α/2 = 0.045.
CI = ( x̄ - z0.045·σ/√n , x̄ + z0.045·σ/√n )
z0.045 ≈ 1.695
CI = ( x̄ ± 1.695·σ/√n )
Part (d): The symmetric interval is usually preferred because, for the same confidence level, it is shorter
(narrower) than the asymmetric interval. The asymmetric form is useful only when unequal tail risks are
specifically desired.
Part (e): A 100(1-α)% lower confidence bound for μ is:
μ ≥ x̄ - zα·σ/√n
So the one-sided lower confidence interval is [ x̄ - zα·σ/√n , ∞ ).
Q2. Required sample size and effect of changing confidence level
Known standard deviation: σ = 12.7. Desired width of the 97% CI = 1.8.
Width W = 2 z(α/2) σ/√n
1.8 = 2(2.1701)(12.7)/√n
n = 937.73 ≈ 938
So, at least 938 observations are required.
If the confidence level increases to 99% while n stays the same, the width becomes:
W99 = 2(2.5758)(12.7)/√938 ≈ 2.14
Hence precision decreases when the confidence level increases: a higher confidence level gives a wider
interval for the same sample size.
To keep the width at 1.8 at 99% confidence, the required sample size would rise to about 1322.
Q3. Effect of lowering confidence from 99% to 90%
Answer: (a) Shorter and would involve a larger risk of being an incorrect interval estimate.
Reason: Lower confidence means a smaller critical value, so the interval becomes narrower, but the chance of
missing the true mean increases.
Q4. Limits for number of defective accounts
Sample proportion of mistakes: p̂ = 45/600 = 0.075.
Estimated number of defective cases in the 2000 accounts = 2000 × 0.075 = 150
SE(p̂ ) = √[ p̂ (1-p̂ )/n × (N-n)/(N-1) ]
SE(total) = 2000 × SE(p̂ ) ≈ 17.998
95% limits = 150 ± 1.96×17.998 = ( 114.72 , 185.28 )
So the expected number of defective cases is about 115 to 185.
Q5. 95% confidence interval for proportion defective
Sample defect proportion: p̂ = 5/100 = 0.05.
95% CI = 0.05 ± 1.96√[(0.05)(0.95)/100]
95% CI = ( 0.0073 , 0.0927 )
Interpretation: We estimate the true defect proportion to be about 5%, and the margin of error is about 4.28
percentage points.
Q6. 95% confidence interval for true average rainfall
n = 18, sample mean = 38.66 mm, sample standard deviation = 8.47 mm.
t0.975,17 = 2.110
95% CI = 38.66 ± 4.21
95% CI = ( 34.45 , 42.87 ) mm
Q7. 95% confidence interval for mean tire tread depth
Given x̄ = 0.32 inch, s = 0.08 inch, n = 10.
t0.975,9 = 2.262
95% CI = 0.32 ± 0.057
95% CI = ( 0.263 , 0.377 ) inch
Q8. 95% confidence interval for average number of crossings
n = 8, sample mean = 14166.25, sample standard deviation = 1945.51.
t0.975,7 = 2.365
95% CI = 14166.25 ± 1626.49
95% CI = ( 12539.76 , 15792.74 ) crossings
Q9. Maximum size of error with 99% confidence
For a 99% confidence interval of the mean, the margin of error is:
E = t0.995,11 × s/√n = 3.106 × 4.28/√12
E ≈ 3.84 beats per minute
So, with 99% confidence, the maximum error in estimating the true average increase is about 3.84 bpm.
Q10. 95% confidence interval for variance and standard deviation
Given n = 20 and sample standard deviation s = 0.1 fl oz.
df = 19
95% CI for variance: ((n-1)s² / χ²0.975,19 , (n-1)s² / χ²0.025,19)
χ²0.025,19 = 8.907, χ²0.975,19 = 32.852
Variance CI = ( 0.005783 , 0.021333 ) fl oz²
Standard deviation CI = ( 0.076 , 0.146 ) fl oz
Q11. Type I error, Type II error, and power of a test
Type I error (α): Rejecting the null hypothesis H0 when H0 is actually true. The probability of a Type I error
is α, the level of significance.
Type II error (β): Failing to reject H0 when H0 is actually false.
Power of a test: The probability of correctly rejecting a false null hypothesis. Power = 1 - β.
A good test has a small α, a small β, and therefore high power.
Q12. Null hypothesis vs alternative hypothesis
Null hypothesis (H0): The statement of no change, no difference, or no effect. It is the hypothesis tested
directly.
Alternative hypothesis (H1 or Ha): The competing statement that an effect, difference, or change exists.
In hypothesis testing, the decision is made in favor of H0 or against H0 in support of the alternative.
Q13. Radio sales after advertisement campaign
(i) Hypotheses:
H0: μ = 200
H1: μ > 200 (improvement means higher average sales)
(ii) Test at 5% significance level, with x̄ = 216, σ = 50, n = 25.
z = (216 - 200)/(50/√25) = 1.60
Critical value at α = 0.05 (right-tailed) = 1.645
Since 1.60 < 1.645, we fail to reject H0.
Conclusion: There is not enough evidence at the 5% level to say that sales improved.
(iii) Rejection begins when x̄ exceeds:
x̄ crit = 200 + 1.645 × (50/√25) = 216.45
So the sample mean must be greater than about 216.45 for H0 to be rejected.
Q14. Telephone calls: test whether the mean length changed
Hypotheses:
H0: μ = 12.44
H1: μ ≠ 12.44
Test statistic (large-sample approximation):
t ≈ z = (13.71 - 12.44)/(2.65/√150) = 5.87
At α = 0.02, the two-tailed critical value is about ±2.33.
Since 5.87 > 2.33, reject H0. The p-value is about 0.000000, which is far below 0.02.
Conclusion: The mean length of long-distance calls in 2011 is significantly different from 12.44 minutes, and
the sample suggests it is higher.
Q15. Does the modified process increase yield?
Known process mean = 500, σ = 9.6, n = 50, sample mean = 535.
Hypotheses:
H0: μ = 500
H1: μ > 500
z = (535 - 500)/(9.6/√50) = 25.78
At α = 0.05, the critical value is z0.05 = 1.645.
Since 25.78 is much larger than 1.645, reject H0.
P-value ≈ 0.000e+00, effectively 0.
Conclusion: The modified process increases the yield significantly.
Final note
For exam writing, Q1, Q2, Q4, Q5, Q6, Q7, Q8, Q9, Q10, Q13, Q14, and Q15 are the numerical parts. Q11
and Q12 are conceptual definitions. The included graphs show the confidence-interval and hypothesis-test
logic.