Part 1 Paired comparison: Reading vs.
Writing scores (n = 250)
(a) State the hypotheses
Because each student took both exams, we analyse the paired differences
di=readi−writeid_i = \text{read}_i-\text{write}_i.
H0:μd=0(no mean difference between exams)HA:μd≠0(a mean difference exists)\
begin{aligned} H_0 &: \mu_d = 0 \quad\text{(no mean difference between exams)}\\ H_A
&: \mu_d \neq 0 \quad\text{(a mean difference exists)} \end{aligned}
(b) Check the conditions for a paired-tt test
Condition How it is met in this study
Paired data Same 250 students supplied both scores.
Random sample / Students were selected at random; 250 ≪ 10 % of all U.S. seniors
independence ⇒ observations are independent between students.
Normality of the Histogram of dd (read – write) is unimodal and roughly symmetric;
differences with n=250>30n=250>30, the Central Limit Theorem also supports
approximate normality of dˉ\bar d.
All requirements are satisfied.
(c) Test statistic, degrees of freedom, and decision
Given
dˉ=−0.545,sd=8.887,n=250\bar d=-0.545,\quad s_d=8.887,\quad n=250
SEdˉ=sdn=8.887250=0.562SE_{\bar d}= \frac{s_d}{\sqrt n}= \frac{8.887}{\
sqrt{250}}=0.562 t=dˉ−0SEdˉ=−0.5450.562≈−0.97t = \frac{\bar d-0}{SE_{\bar d}} =\frac{-
0.545}{0.562} \approx -0.97
Degrees of freedom for a paired-tt is n−1=249n-1 = 249.
The problem supplies p = 0.39 (two-tailed).
Decision at α=0.05
Because p=0.39>0.05p=0.39 > 0.05, fail to reject H0H_0.
There is not convincing evidence of a difference in average reading and writing scores.
(d) Possible error
We did not reject H0H_0; the only error we could now make is a Type II error: concluding
“no difference” when, in truth, a real mean difference exists.
Context: we might be overlooking a genuine gap between students’ reading and writing
performances.
(e) Would a confidence interval include 0?
Yes. Failing to reject H0H_0 implies the 95 % CI for μd\mu_d should contain 0.
(Indeed, a quick CI uses dˉ±t0.025,249\* SE≈−0.545±1.97(0.562)\bar d \pm t^\
*_{0.025,249}\,SE \approx -0.545 \pm 1.97(0.562), i.e. from about –1.65 to +0.56, which
includes 0.)
Part 2 – Manual vs Automatic City MPG (EPA sample, n₁ = n₂ = 26)
Transmission Mean MPG SD n
Automatic 16.12 3.58 26
Manual 19.85 4.51 26
1) State the hypotheses
H0:μmanual−μauto=0(no difference in average city MPG)HA:μmanual−μauto≠0(a difference
exists)\begin{aligned} H_0 &: \mu_{\text{manual}}-\mu_{\text{auto}} = 0 \quad\text{(no
difference in average city MPG)}\\ H_A &: \mu_{\text{manual}}-\mu_{\text{auto}} \neq 0 \
quad\text{(a difference exists)} \end{aligned}
(If your course wants a one-sided test in favour of manuals, change HAH_A to “ > 0”.)
2) Calculate the tt-statistic (Welch two-sample tt)
Difference of sample means: xˉ M−xˉ
A=19.85−16.12=3.73SE=sA2nA+sM2nM=3.58226+4.51226≈1.13t=3.731.13≈3.30\
begin{aligned} \text{Difference of sample means} &: \; \bar x_{\!M}-\bar x_{\!A}=19.85-
16.12 = 3.73 \\[4pt] SE &= \sqrt{\frac{s_A^{2}}{n_A}+\frac{s_M^{2}}{n_M}} = \sqrt{\
frac{3.58^{2}}{26}+\frac{4.51^{2}}{26}} \approx 1.13 \\[4pt] t &= \frac{3.73}{1.13} \
approx \boxed{3.30} \end{aligned}
3) Degrees of freedom (Welch–Satterthwaite)
df = (sA2nA+sM2nM) 2(sA2/nA)2nA−1+(sM2/nM)2nM−1 ≈ 48 (rounded)df \;=\; \frac{\
left(\dfrac{s_A^{2}}{n_A}+\dfrac{s_M^{2}}{n_M}\right)^{\!2}}
{\dfrac{\bigl(s_A^{2}/n_A\bigr)^{2}}{n_A-1}+ \dfrac{\bigl(s_M^{2}/n_M\bigr)^{2}}
{n_M-1}} \;\approx\; \boxed{48\text{ (rounded)}}
(A calculator gives 47.7, which we round to 48.)
4) Conclusion using p=0.0029p = 0.0029
At the usual significance level α=0.05\alpha = 0.05 (and even α=0.01\alpha = 0.01):
p=0.0029<αp = 0.0029 < \alpha
Reject H0H_0.
Interpretation: The data provide strong evidence that cars with manual transmissions have a
different—specifically higher—average city fuel efficiency than cars with automatic
transmissions for the model year studied.
That covers all four requested items. Let me know if you’d like the confidence-interval
calculation or any additional context!
