Part 1: Reading vs.
Writing Scores
(a) State the Hypotheses
Let μd be the true mean difference between reading and writing scores (read - write).
Null Hypothesis H o : μd =0 (no difference in average scores)
Alternative Hypothesis H A : μd ≠ 0 (there is a difference)
(b) Check Conditions for the Test
Randomness: The sample is randomly selected.
Independence: The sample of 250 is <10% of all students, so observations are
likely independent.
Normality: With a large sample (n = 250), the Central Limit Theorem ensures the
sampling distribution is approximately normal, even if the population is not.
(c) Hypothesis Test Calculations
Given:
x d =−0.545
sd =8.887
n=250
Test Statistic (T-score):
x d −0 −0.545
t= = ≈−0.97
sd / √ n 8.887 / √ 250
Degrees of Freedom:
Since it's a one-sample t-test for differences:
df =n−1=249
Conclusion (with p-value = 0.39):
Since the p-value (0.39) > 0.05, fail to reject H o .
Conclusion: There is not enough evidence to suggest a significant difference
between average reading and writing scores.
(d) Type of Error
Failing to reject H o when there is actually a difference is a Type II error. In this context,
we might have missed a real difference in average reading vs. writing scores.
(e) Confidence Interval Interpretation
Since we failed to reject H o , a 95% confidence interval would likely include 0,
supporting the conclusion that there's no significant difference.
Part 2: Automatic vs. Manual Transmission (City MPG)
Given:
x 1=16.12 , s 1=3.58 , n1=26( Automatic)
x 2=19.85 , s 2=4.51 , n2=26(Manual)
(1) State the Hypotheses
Let μ1=meancity MPG for automatic
Let μ2=meancity MPG for manual
H o : μ1=μ2
H a : μ1 ≠ μ2
(2) Calculate the T-statistic
x 1−x 2 16.12−19.85
t= = ≈−3.30
√ √
2 2 2 2
s1 s2 3.58 4.51
+ +
n1 n 2 26 26
(3) Degrees of Freedom (Welch-Satterthwaite Formula)
2 2 2
s1 s2
( + )
n1 n 2
df = ≈ 46.8 ≈ 47
s 21 2 s 22 2
( ) ( )
n1 n2
+
n1−1 n2−1
(4) Conclusion (p-value = 0.0029)
Since p = 0.0029 < 0.05, reject H o .
Conclusion: There is strong evidence that the average city MPG differs between
automatic and manual transmission cars - manual cars appear to be more fuel-
efficient.
Discussion Post: Paired vs. Unpaired Data and T-Distribution
In statistical analysis, the use of the t-distribution is vital when working with small
sample sizes and unknown population standard deviations. It is especially useful when
analyzing paired and unpaired data. To illustrate these concepts, I will provide one
example each of paired and unpaired data from the field of healthcare.
Paired Data Example:
Consider a study assessing the effectiveness of a new blood pressure medication. The
blood pressure of 30 patients is recorded before and after taking the medication for six
weeks. Since measurements are taken on the same subjects before and after the
intervention, this is an example of paired data. The data points are dependent because
each patient's before and after values are linked. The correct statistical test for analyzing
this dataset would be the one-sample t-procedure applied to the differences (post-
treatment – pre-treatment), as we are interested in whether the average difference in
blood pressure is significantly different from zero (Moore et al., 2021).
Unpaired Data Example:
Now consider a study comparing the average cholesterol levels of two independent
groups: 40 males and 40 females. Since the individuals in one group are not related or
matched with those in the other, this is an example of unpaired (independent) data. The
appropriate test here would be the two-sample t-procedure, as we are comparing the
means of two independent groups to determine if a statistically significant difference
exists.
Distinguishing Paired vs. Unpaired Data:
The key distinction between these types lies in the relationship between observations.
Paired data involves dependent samples, typically arising from measurements taken on
the same subject under different conditions or at different times. In contrast, unpaired
data involves independent samples, with no logical pairing between observations.
Using the correct t-procedure is critical to avoid erroneous conclusions. Applying a two-
sample t-test to paired data or vice versa would violate the assumptions of the test and
compromise the validity of the results.
References:
Moore, D. S., Notz, W. I., & Fligner, M. A. (2021). The Basic Practice of Statistics (9th
ed.). W.H. Freeman.
