STATISTICS

Complete Exam Study Guide

All Practice Questions — Fully Solved with Explanations

Formula Reference & Cheat Sheet Guidance

SECTION TOPIC

PART 1 Exam Regulations & Cheat Sheet Guide

PART 2 Topic Overview & Key Formulas

PART 3 Pre-Exam Practice (28 May) — All Solutions

PART 4 Ten Exercises — All Solutions

PART 1 — Exam Regulations & Cheat Sheet Guidance

What You Are Allowed to Bring

Item Details

Calculator Non-programmable only

Formula sheet 1 A4 sheet, front & back, handwritten or typed — YOUR name on it

Statistical tables 1 separate sheet for standard tables (Z, t, chi-square)

Cheat Sheet — Priority Formulas

Every formula below is used in at least one exam exercise:

# Formula Used in

1 Sample mean Ex 1,2,4,5,6,7,10

2 Sample variance (n-1) Ex 1,4,6

3 Sample std dev CI & t-test

4 95% CI — sigma unknown (t) Ex 1,6,7,10

5 95% CI — sigma known (Z) Ex 5,7

6 One-sample t-statistic Ex 2,6

7 Pooled SD (two groups) Ex 10

8 CI two means Ex 10

9 Sensitivity Ex 3,5

10 Specificity Ex 3,5

11 PPV (Bayes) Ex 3,5

12 NPV (Bayes) Ex 5

13 Chi-square statistic Ex 4,6

14 Expected cell count Ex 4,6

15 Covariance Ex 4 pre-exam

16 Pearson r Regression

17 Regression slope b1 Ex 4 pre-exam

18 Poisson probability Ex 6b

19 Power of test Ex 5 pre-exam

20 Sample size formula Ex 9

21 Proportion CI Ex 8

Statistical Tables to include on your separate sheet: t-distribution table (df 1–30+, alpha 0.025 & 0.05), standard
normal Z-table, chi-square table (df 1–10).

Exam Rules Reminders

• All final answers must have exactly 4 significant figures.
• Always show full calculation steps — methodology earns marks.
• Section A (MCQ): 15 min. Section B (4 exercises): 75 min. Total: 32 marks.
• Heuristic explanations alone are NOT sufficient — the math must be shown.
PART 2 — Topic Overview & Key Formulas

1. Descriptive Statistics
The starting point of every exercise. Always compute these first.

Key: always divide by (n-1) for sample variance, NOT n. If the problem gives you sigma (population std dev),
that is different — use it directly.

2. Confidence Intervals
When population sigma is KNOWN — use Z:

When sigma is UNKNOWN (most cases) — use t:

 Interpretation: 'We are 95% confident the true mean lies between [lower, upper].' This does NOT mean 95% of
data values fall here.

3. Hypothesis Testing — One-Sample t-Test

Right-tailed: reject H0 if t > t_crit. Left-tailed: reject H0 if t < −t_crit. Two-tailed: reject H0 if |t| > t_crit.

4. Diagnostic Testing
Disease + Disease − Total

Test + TP FP TP+FP

Test − FN TN FN+TN

Total TP+FN FP+TN N

5. Correlation & Linear Regression

6. Chi-Square Test of Independence

7. Poisson Distribution
8. Power of a Test

9. Sample Size

10. CI for a Proportion

Normal approx valid when: n·p-hat ≥ 5 AND n·(1−p-hat) ≥ 5

11. Two Independent Means — Pooled

PART 3 — Pre-Exam Practice (28 May 2026) — Full Solutions

SECTION A — Multiple Choice

Q1. Which of the following is a continuous variable?
A. Blood type B. Number of hospital admissions C. Serum cholesterol level D. Stage of cancer (I–IV)
ANSWER: C — Serum cholesterol level
Blood type (A) is nominal categorical. Number of admissions (B) is discrete (whole numbers). Cancer
stage (D) is ordinal categorical. Serum cholesterol (C) is continuous — it can take any decimal value within
a range (e.g., 185.4 mg/dL).

Q2. Which is NOT an assumption of the one-sample t-test?

A. Normality of population B. Independence of observations C. Known population variance D. Continuous
measurement scale
ANSWER: C — Known population variance
The t-test was invented precisely for when sigma is UNKNOWN — we substitute s. The three real
assumptions are: normality (A), independence (B), continuous scale (D).

SECTION B — Open Problems

Exercise 3 — Diagnostic Testing (6 marks)

1,000 subjects. 200 truly diseased, 800 healthy. 170 true positives, 40 false positives.

Part (a) — 2×2 Contingency Table

TP = 170, FN = 200 − 170 = 30, FP = 40, TN = 800 − 40 = 760

Disease + Disease − Total

Test + 170 (TP) 40 (FP) 210

Test − 30 (FN) 760 (TN) 790

Total 200 800 1000

Part (b) — Sensitivity and Specificity

Sensitivity = 0.8500 (85.00%), Specificity = 0.9500 (95.00%)

Part (c) — PPV

 PPV = 0.8095 (80.95%)
Interpretation: A patient who tests positive has an 80.95% probability of truly being infected. About 1 in 5
positive results is a false alarm in this sample (20% prevalence).

Exercise 4 — Correlation & Linear Regression (6 marks)

Age X (years) and Height Y (cm) for 10 individuals.

Person X Y xi−x■ yi−■ (xi−x■)(yi−■) (xi−x■)² (yi−■)²

1 18 165 −8.6 −9.7 83.42 73.96 94.09

2 20 168 −6.6 −6.7 44.22 43.56 44.89

3 22 170 −4.6 −4.7 21.62 21.16 22.09

4 25 172 −1.6 −2.7 4.32 2.56 7.29

5 27 173 0.4 −1.7 −0.68 0.16 2.89

6 30 175 3.4 0.3 1.02 11.56 0.09

7 32 178 5.4 3.3 17.82 29.16 10.89

8 35 180 8.4 5.3 44.52 70.56 28.09

9 37 181 10.4 6.3 65.52 108.16 39.69

10 40 183 13.4 8.3 111.22 179.56 68.89

Σ 286 1745 0 0 392.00 540.40 319.00

Step 1 — Means

x■ = 28.60 years, ■ = 174.5 cm

Step 2 — Sample Variances and Standard Deviations

sX² = 60.04, sX = 7.749 years | sY² = 35.44, sY = 5.953 cm

Step 3 — Covariance
 Cov(X,Y) = 43.56

Step 4 — Pearson r

r = 0.9443 — very strong positive linear correlation.

Step 5 — Regression Line

■ = 153.8 + 0.7255 X
For each extra year of age, height increases by ~0.73 cm. The intercept 153.8 is mathematically required
but has no physiological meaning at age 0.

Exercise 5 — Power of a Test (6 marks)

X ~ N(µ, σ²=36), so σ=6, n=25. H0: µ=20 vs H1: µ=24 (right-tailed). Critical value given: x■c = 21.97.

Step 1 — Verify critical value

Step 2 — Power under H1: µ1 = 24

Power = 0.9549 (≈95.49%)

If the true mean is 24, there is a 95.49% chance the test correctly rejects H0. This is very high power.

Exercise 6(a) — t-Test: Impurities (6 marks)

Historical mean µ0 = 5.40 g. Data (n=10): 4.9, 5.1, 4.7, 5.0, 4.8, 4.9, 5.2, 4.6, 4.8, 4.9

Part (a) — Sample mean and std dev

x■ = 4.890 g, s = 0.1792 g

Part (b) — 95% CI (df = 9, t0.025,9 = 2.262)

95% CI: (4.762 , 5.018) grams

Part (c) — Hypotheses

Part (d) — Test at α = 1% (t0.01,9 = −2.821)

t = −9.000 < −2.821 → REJECT H0 at the 1% level.

Conclusion: Very strong evidence that the new purification method significantly reduces average impurities
below the historical 5.40 g per batch.

Exercise 6(b) — Poisson Distribution: Emergency Calls (6 marks)

X ~ Poisson(λ = 3) per hour.

Part 1 — P(X = 4)

P(X = 4) = 0.1680

Part 2 — P(X ≤ 2)

P(X ≤ 2) = 0.4232

Part 3 — P(3 ≤ X ≤ 6)

P(3 ≤ X ≤ 6) = 0.5432
PART 4 — Ten Exercises — Full Solutions

Exercise 1 — Descriptive Statistics & Normality (6 marks)

Data: 12.1, 11.8, 12.4, 12.0, 11.9, 12.3, 12.2, 12.0 (n=8)

Part (a) — Mean and Standard Deviation

x■ = 12.09 mg/mL, s = 0.2031 mg/mL

Part (b) — 95% CI (df=7, t0.025,7 = 2.365)

95% CI: (11.92 , 12.26) mg/mL

Part (c) — Interpretation

We are 95% confident that the true mean concentration lies between 11.92 and 12.26 mg/mL. The narrow
interval suggests consistent manufacturing quality.

Exercise 2 — One-Sample t-Test (6 marks)

n=15, x■=19.2 g, s=1.8 g. Test if mean protein < 20 g.

Part (a) — Hypotheses

Part (b) — Test Statistic

t = −1.721

Part (c) — Decision at α=0.05 (df=14, t_crit=−1.761)

t = −1.721 > −1.761 → FAIL TO REJECT H0 at α = 0.05.
Conclusion: Insufficient evidence that mean protein content is less than 20 g. The 0.8 g shortfall could be
due to sampling variability.

Exercise 3 — Diagnostic Testing with Low Prevalence (6 marks)

500 subjects: 120 infected, 380 healthy. 108 TP, 25 FP.

Part (a) — 2×2 Table

Infected + Healthy − Total

Test + 108 (TP) 25 (FP) 133

Test − 12 (FN) 355 (TN) 367

Total 120 380 500

Part (b) — Sensitivity and Specificity

Sensitivity = 0.9000, Specificity = 0.9342

Part (c) — PPV with 8% prevalence (Bayes' Theorem)

PPV = 0.5433 (54.33%)

At only 8% prevalence, half of all positive results are false positives — low prevalence dramatically
reduces PPV even for a good test.

Exercise 4 — Chi-Square Test of Independence (6 marks)

Smoking vs Chronic Cough, n=300.

Chronic Cough No Cough Total

Smokers 54 66 120

Non-smokers 36 144 180

Total 90 210 300

Part (a) — Expected Counts

Part (b) — Chi-Square Statistic

 χ² = 21.43

Part (c) — Decision (df=1, χ²_crit = 3.841 at α=0.05)

21.43 >> 3.841 → REJECT H0. Smoking and chronic cough are significantly associated.

Exercise 5 — Full Bayes Analysis (6 marks)

400 infected: 384 TP, 16 FN. 600 non-infected: 12 FP, 588 TN. Ward prevalence = 2%.

Infected + Non-infected − Total

Test + 384 (TP) 12 (FP) 396

Test − 16 (FN) 588 (TN) 604

Total 400 600 1000

Part (b) — Sensitivity and Specificity

Sensitivity = 0.9600, Specificity = 0.9800

Part (c) — PPV at P(I)=0.02

PPV = 0.4948 (49.48%)

Part (d) — NPV at P(I)=0.02

NPV = 0.9992 (99.92%)

Part (e) — Interpretation of low prevalence

At 2% prevalence, even with Sensitivity=96% and Specificity=98%, the PPV is only ~49.5% — roughly half
of positives are false alarms. The large pool of healthy patients generates more false positives than the
small infected group generates true positives. However, a negative result is extremely reliable
(NPV=99.92%). Clinical takeaway: confirm all positives before prescribing high-potency antibiotics.

Exercise 6 — Chi-Square: Dosage vs Muscle Pain (6 marks)

Muscle Pain No Muscle Pain Total

Low Dose 15 85 100

High Dose 35 65 100

 Muscle Pain No Muscle Pain Total

Total 50 150 200

H0: Dosage and muscle pain are independent. H1: They are associated.

Part (b) — Expected Frequencies

Part (c) — Chi-Square Statistic

χ² = 10.67 > 6.635 (χ²_crit at α=0.01, df=1) → REJECT H0 even at 1%.

Conclusion: Strong evidence of a significant association between dosage level and muscle pain.
High-dose patients (35%) report pain at more than twice the rate of low-dose patients (15%).

Exercise 7 — CI for Mean, Sigma Known (6 marks)

n=36, x■=118 s, σ=12 s.

Part (a) — 95% CI

95% CI: (114.1 , 121.9) seconds

Part (b) — Regulatory interpretation

Based on 36 tablets, we are 95% confident the true mean dissolution time lies between 114.1 and 121.9
seconds, indicating a consistent profile within approximately ±4 seconds.

Exercise 8 — CI for Proportion (6 marks)

n=1200 newborns, 18 positive.

Part (a) — Sample proportion

p■ = 0.01500 (1.500%)

Part (b) — 95% CI

95% CI: (0.008122 , 0.02188)

Part (c) — Validity of normal approximation

Both conditions met — normal approximation is appropriate here. For very small p■, exact binomial
methods (Wilson/Clopper-Pearson) would be preferred in clinical settings.

Exercise 9 — Sample Size (6 marks)

Margin E ≤ 3 mmHg, 95% confidence, σ = 15 mmHg.

Part (a) — Minimum n

Minimum sample size: n = 97

Part (b) — Effect of reducing margin to ±2 mmHg

Sample size increases from 97 to 217 (~2.24× more). Since n ∝ 1/E², reducing the margin of error requires
disproportionately larger samples.

Exercise 10 — CI for Difference Between Two Means (6 marks)

Group A: n1=40, x■1=92, s1=10. Group B: n2=35, x■2=87, s2=9.

Part (a) — Pooled Standard Deviation

sp = 9.548 mg/dL

Part (b) — 95% CI (df=73, t0.025,73 ≈ 1.993)

95% CI for (µ1 − µ2): (0.5960 , 9.404) mg/dL

Part (c) — Interpretation

We are 95% confident that the standard-diet group has a mean fasting glucose between 0.596 and 9.404
mg/dL higher than the high-fiber group. Since the entire CI is positive (above zero), this is statistically
significant evidence that the high-fiber diet is associated with meaningfully lower glucose levels.

END OF STUDY GUIDE — Good luck on your exam!