Background

This exercise analyses weight loss data from Dataset B (Diets). The goal is to calculate summary measures for Diet B and compare them with Diet A to assess the relative effectiveness of the two weight reducing diets.

Excel analysis showing summary statistics for Diet A and Diet B

Interpretation

The results clearly indicate that Diet A is more effective than Diet B for weight reduction:

Central Tendency: The mean weight loss for Diet A (5.341 kg) is approximately 1.63 kg higher than Diet B (3.710 kg). Similarly, the median for Diet A (5.642 kg) exceeds Diet B's median (3.745 kg) by about 1.9 kg. Both measures of central tendency consistently favour Diet A.

Variability: The standard deviations are similar (Diet A: 2.536 kg, Diet B: 2.769 kg), indicating comparable spread in weight loss outcomes for both diets. The IQRs are also similar, suggesting the middle 50% of participants experienced similarly variable results.

Practical Significance: Both diets produce positive average weight loss, indicating both are effective to some degree. However, Diet A produces approximately 44% more weight loss on average, which represents a clinically meaningful difference. This suggests Diet A should be recommended over Diet B for individuals seeking to maximise weight loss.

Background

This exercise analyses brand preference data from Dataset D (Brandprefs). The goal is to calculate frequencies and percentage frequencies for Area 2 respondents and compare brand preference patterns between the two demographic areas.

Frequency and percentage analysis for brand preferences across two areas

Interpretation

The brand preference patterns differ notably between the two demographic areas:

Brand A: Shows higher preference in Area 2 (21.1%) compared to Area 1 (15.7%), a difference of 5.4 percentage points.

Brand B: Also shows substantially higher preference in Area 2 (33.3%) compared to Area 1 (24.3%), a difference of 9 percentage points.

Other Brands: The preference for competitor brands is notably lower in Area 2 (45.6%) compared to Area 1 (60.0%), a difference of 14.4 percentage points.

Marketing Implications: These findings suggest that the manufacturer's brands (A and B) have stronger market penetration in Area 2, with over half (54.4%) of respondents preferring either Brand A or B. In contrast, Area 1 shows weaker brand loyalty with only 40% preferring the manufacturer's brands. This could inform targeted marketing strategies. Area 1 may require more intensive promotional efforts to increase brand awareness and loyalty.

Background

This exercise analyses filtration data from Dataset G. Each batch was split and filtered using two different agents, making a paired (related) samples t-test appropriate. The goal is to test whether the population mean impurity differs between the two filtration agents.

Paired t-test analysis for filtration data comparing two agents

Interpretation

Statistical Conclusion: The two-tailed p-value (0.0075) is less than 0.05, so we reject the null hypothesis at the 5% significance level. There is significant evidence that the population mean impurity differs between the two filtration agents.

Practical Interpretation: Agent 1 produces a mean impurity of 8.250 parts per 1000, which is 0.433 parts per 1000 lower than Agent 2's mean of 8.683. This difference is statistically significant (p = 0.0075, which is significant at the 1% level).

Recommendation: Since lower impurity indicates better filtration performance, Agent 1 appears to be the more effective filtration agent and should be preferred for applications where minimising impurity is important.

Background

Building on Exercise 7.2.4, we now conduct a one-tailed test to determine whether Filter Agent 1 is specifically more effective (produces lower impurity) than Agent 2.

Interpretation

Statistical Conclusion: The one-tailed p-value (0.0038) is less than 0.01, so we reject the null hypothesis at the 1% significance level. There is strong evidence that Filter Agent 1 is more effective at removing impurities than Agent 2.

Comparison with Two-Tailed Test: The one-tailed test provides stronger evidence (p = 0.0038 vs p = 0.0075) because we had a directional hypothesis. The one-tailed p-value is exactly half the two-tailed p-value when the sample mean difference is in the predicted direction.

Conclusion: There is statistically significant evidence at the 1% level that Filter Agent 1 produces lower impurity levels, making it the preferred choice for filtration applications.

Background

This exercise analyses bank cardholder data from Dataset C (Superplus). The goal is to test whether the population mean income for male cardholders exceeds that of female cardholders. Since the male and female samples are independent, an independent samples t-test is appropriate.

Independent samples t-test comparing male and female cardholder incomes

Step 1: F-test for Equality of Variances

Step 2: Independent Samples t-Test (One-tailed)

Interpretation

Statistical Conclusion: The one-tailed p-value (0.0007) is less than 0.01, so we reject the null hypothesis at the 1% significance level. There is strong evidence that the population mean income for male Superplus Diamond cardholders exceeds that of female cardholders.

Practical Interpretation: Male cardholders have a mean income of £52,913, which is £8,680 higher than the female mean income of £44,233. This represents a 19.6% higher income for males on average.

Assumptions and Validation:

1. Independence: The samples are independently drawn. Each cardholder appears in only one group.
2. Normality: Income should be approximately normally distributed. With n=60 in each group, the Central Limit Theorem supports the validity of the t-test even if distributions are slightly non-normal. Validation could be performed using normal probability plots (Q-Q plots) or histograms for each group.
3. Equal Variances: The F-test (p ≈ 0.44) confirms this assumption is reasonable.
4. Random Sampling: We assume cardholders were randomly selected from the Superplus Diamond cardholder population.