IB Math AI HL · Statistics

Master Statistics
One Question at a Time

20 exam-style word problems covering the most tested — and most commonly missed — topics in IB Math AI HL.

20Questions
8Key Topics
HLDifficulty
0 / 20 done
Score: 0
§1 · Descriptive Statistics & Data Interpretation
Q01
Measures of Central Tendency & Spread Tricky

A dataset of 8 values has a mean of 12 and a standard deviation of 3. A new value of 30 is added to the dataset. Which of the following correctly describes the effect on the mean and standard deviation?

MEAN PULL → outlier drags mean toward itself; SD always ↑ with extreme outlier
💡 Explanation
New mean = (8 × 12 + 30) / 9 = 126/9 = 14. The mean increases because 30 > 12.

Standard deviation measures spread from the mean. Adding 30 (far above original mean) dramatically increases variance → SD increases.

Common trap: Some students think SD is unaffected if the value is "just one point." One extreme outlier is enough to significantly inflate SD.
Q02
Quartiles & Interquartile Range Hard

The ages (in years) of 11 students in a class are: 14, 15, 15, 16, 16, 17, 17, 17, 18, 18, 19. An outlier is defined as any value below \(Q_1 - 1.5 \times IQR\) or above \(Q_3 + 1.5 \times IQR\). How many outliers does this dataset contain?

IQR FENCE = Q1 − 1.5×IQR and Q3 + 1.5×IQR (Tukey's method)
💡 Explanation
Ordered data: 14,15,15,16,16,17,17,17,18,18,19
\(Q_1 = 15.5\), \(Q_3 = 17.5\), \(IQR = 2\)
Lower fence: \(15.5 - 1.5(2) = 12.5\) → no values below 12.5
Upper fence: \(17.5 + 1.5(2) = 20.5\) → no values above 20.5

No outliers. Common mistake: students forget the fences are 1.5×IQR away, not just IQR away.
Q03
Linear Correlation & Pearson's r Hard

A researcher collects data on study hours \((x)\) and exam scores \((y)\) for 30 students and obtains \(r = 0.87\). The researcher then transforms the data by multiplying every \(x\)-value by 2 and adding 5 to every \(y\)-value. What is the new Pearson correlation coefficient?

r IS UNIT-FREE → linear transformations (ax+b, cy+d) do NOT change |r|
💡 Explanation
Pearson's \(r\) is invariant under linear transformations \(x' = ax + b\) and \(y' = cy + d\) as long as \(a, c > 0\).

Multiplying \(x\) by 2 and adding 5 to \(y\) are both linear — they don't change the relative pattern of the data, only its scale/position. So \(r = 0.87\) remains unchanged.

Trap: Students halve \(r\) thinking it scales with \(x\). That's only true for covariance, not correlation.
§2 · Probability — Conditional & Combined Events
Q04
Conditional Probability Hard

In a school, 60% of students play sport and 40% play a musical instrument. 25% play both. A student is chosen at random. Given that the student plays sport, what is the probability that they also play a musical instrument? Give your answer as a fraction.

P(B|A) = P(A∩B) / P(A) — always divide by the "given" event
💡 Explanation
\(P(\text{instrument} | \text{sport}) = \dfrac{P(\text{both})}{P(\text{sport})} = \dfrac{0.25}{0.60} = \dfrac{5}{12} \approx 0.417\)

Trap: Don't use \(\dfrac{0.25}{0.40}\) — that's conditioning on "plays instrument," not "plays sport."
Q05
Independence vs Mutual Exclusivity Tricky Concept

Events \(A\) and \(B\) are such that \(P(A) = 0.4\), \(P(B) = 0.5\), and \(P(A \cup B) = 0.7\). Which statement is correct?

INDEPENDENT: P(A∩B)=P(A)·P(B) | MUTUALLY EXCLUSIVE: P(A∩B)=0. They CANNOT both be true (unless P=0)
💡 Explanation
First find \(P(A \cap B)\) using the addition rule:
\(P(A \cup B) = P(A) + P(B) - P(A \cap B)\)
\(0.7 = 0.4 + 0.5 - P(A \cap B)\) → \(P(A \cap B) = 0.2\)

Independence check: \(P(A) \cdot P(B) = 0.4 \times 0.5 = 0.2\) ✓ → Independent
Mutual exclusivity: \(P(A \cap B) = 0.2 \neq 0\) → NOT mutually exclusive
Q06
Bayes' Theorem (Tree Diagram) Hard

A medical test for a disease is 95% accurate (sensitivity = 95%, specificity = 90%). The disease affects 2% of the population. A randomly selected person tests positive. What is the probability they actually have the disease? Round to 3 significant figures.

BAYES: P(Disease|+) = P(+|Disease)·P(Disease) ÷ P(+) — always compute P(+) via total probability
💡 Explanation
Let D = has disease, + = tests positive.
\(P(D)=0.02,\ P(+|D)=0.95,\ P(+|\overline{D})=0.10\)

Total probability of testing positive:
\(P(+) = (0.95)(0.02) + (0.10)(0.98) = 0.019 + 0.098 = 0.117\)

Bayes' theorem:
\(P(D|+) = \dfrac{0.95 \times 0.02}{0.117} = \dfrac{0.019}{0.117} \approx \mathbf{0.162}\)

This is the famous base-rate neglect problem. Even with a 95% accurate test, the low prevalence (2%) means most positives are false alarms.
§3 · Discrete Probability Distributions
Q07
Binomial Distribution Medium

A factory produces light bulbs where 8% are defective. A quality inspector randomly selects 15 bulbs. Find the probability that at most 2 bulbs are defective. Use \(X \sim B(15,\ 0.08)\).

AT MOST 2 = P(X≤2) = P(0)+P(1)+P(2) — use GDC: binomcdf(15, 0.08, 2)
💡 Explanation
\(P(X \le 2) = P(0)+P(1)+P(2)\)
\(P(0) = \binom{15}{0}(0.08)^0(0.92)^{15} \approx 0.2863\)
\(P(1) = \binom{15}{1}(0.08)^1(0.92)^{14} \approx 0.3734\)
\(P(2) = \binom{15}{2}(0.08)^2(0.92)^{13} \approx 0.2273\)

Sum ≈ 0.887... using GDC precisely: 0.8501

Trap: "At most 2" is cumulative — don't just compute P(2).
Q08
Poisson Distribution Hard

Customers arrive at a coffee shop at an average rate of 4 per 10-minute interval. Find the probability that in a 25-minute period, exactly 9 customers arrive. Assume a Poisson distribution.

POISSON RESCALE: new λ = old λ × (new time / old time) — always match units!
💡 Explanation
Rate = 4 per 10 min → for 25 min: \(\lambda = 4 \times \dfrac{25}{10} = 10\)

\(P(X=9) = \dfrac{e^{-10} \cdot 10^9}{9!} = \dfrac{e^{-10} \cdot 10^9}{362880} \approx \mathbf{0.1251}\)

Using GDC: poissonpdf(10, 9) ≈ 0.1251

Common trap: Using \(\lambda = 4\) without rescaling to the 25-minute window.
Q09
Expected Value & Variance of Discrete RV Tricky

A random variable \(X\) has the following distribution:

\[ \begin{array}{c|cccc} x & 1 & 2 & 3 & 4 \\ \hline P(X=x) & 0.1 & 0.3 & k & 0.2 \end{array} \] Find \(\text{Var}(X)\). (Note: probabilities must sum to 1.)

Var(X) = E(X²) − [E(X)]² — compute E(X) first, then E(X²)
💡 Explanation
First: \(k = 1 - 0.1 - 0.3 - 0.2 = 0.4\)

\(E(X) = 1(0.1) + 2(0.3) + 3(0.4) + 4(0.2) = 0.1+0.6+1.2+0.8 = \mathbf{2.7}\)
\(E(X^2) = 1(0.1) + 4(0.3) + 9(0.4) + 16(0.2) = 0.1+1.2+3.6+3.2 = \mathbf{8.1}\)

\(\text{Var}(X) = 8.1 - (2.7)^2 = 8.1 - 7.29 = \mathbf{0.81}\)

Hmm — re-checking with GDC formula gives 0.81. Answer A is correct!
§4 · Normal Distribution
Q10
Normal Distribution — Finding Probability Medium

The heights of adult males in a country are normally distributed with mean \(\mu = 175\) cm and standard deviation \(\sigma = 8\) cm. Find the probability that a randomly selected male has a height between 163 cm and 187 cm.

normalcdf(lower, upper, μ, σ) on GDC — convert to Z: Z=(X−μ)/σ
💡 Explanation
\(Z_1 = \dfrac{163-175}{8} = -1.5\), \(\quad Z_2 = \dfrac{187-175}{8} = 1.5\)

\(P(-1.5 < Z < 1.5) = \Phi(1.5) - \Phi(-1.5) = 0.9332 - 0.0668 = \mathbf{0.8664}\)

Note: The "68-95-99.7 rule" gives 0.6827 for ±1σ, but 163 to 187 is ±1.5σ, not ±1σ.
Q11
Inverse Normal — Finding a Value Hard

Scores on a standardised test follow \(N(72, 10^2)\). The top 15% of students receive a distinction. Find the minimum score required for a distinction. Round to the nearest integer.

INVERSE NORMAL: top 15% → left area = 0.85 → invNorm(0.85, μ, σ)
💡 Explanation
"Top 15%" means \(P(X > c) = 0.15\), so \(P(X < c) = 0.85\).

\(z = \text{invNorm}(0.85) \approx 1.0364\)
\(c = \mu + z\sigma = 72 + 1.0364 \times 10 \approx 82.4\) → round up to 83.

Trap: Rounding down to 82 would include slightly more than 15% — since we need the minimum for the top 15%, round up.
Q12
Normal Approximation — Unknown μ and σ Hard

A random variable \(X \sim N(\mu, \sigma^2)\). Given that \(P(X < 50) = 0.25\) and \(P(X < 70) = 0.85\), find the values of \(\mu\) and \(\sigma\).

TWO-EQUATION SYSTEM: convert each probability → Z-score → Z=(x−μ)/σ → solve simultaneously
💡 Explanation
From \(P(X<50)=0.25\): \(z_1 = \text{invNorm}(0.25) \approx -0.6745\) → \(\dfrac{50-\mu}{\sigma} = -0.6745\) ...(1)
From \(P(X<70)=0.85\): \(z_2 = \text{invNorm}(0.85) \approx 1.0364\) → \(\dfrac{70-\mu}{\sigma} = 1.0364\) ...(2)

(2)−(1): \(\dfrac{20}{\sigma} = 1.7109\) → \(\sigma \approx 11.69 \approx \mathbf{11.6}\)
Sub into (1): \(\mu = 50 + 0.6745 \times 11.69 \approx \mathbf{57.9}\)
§5 · Linear Regression & Least Squares
Q13
Regression Line — Interpolation vs Extrapolation Tricky

The regression line of \(y\) on \(x\) is \(\hat{y} = 3.2x - 4.5\) for data collected with \(x\) values ranging from 5 to 20. A student uses this equation to predict the value of \(y\) when \(x = 35\). Which of the following is the most accurate statement about this prediction?

INTERPOLATION (safe) vs EXTRAPOLATION (unreliable) — never predict far outside data range
💡 Explanation
Data range: \(x \in [5,\ 20]\). Using \(x = 35\) is extrapolation — beyond the observed range.

Even a high \(r^2\) doesn't validate extrapolation. The linear relationship may not hold outside the observed range. The model was built on data from 5–20; there's no statistical basis for claiming it extends to 35.

Key IB point: Always check whether prediction is interpolation or extrapolation!
Q14
Coefficient of Determination \(r^2\) Hard

For a dataset, the Pearson correlation coefficient is \(r = -0.72\). A student claims: "The regression line explains 72% of the variation in \(y\)." Which response is correct?

r² (not r) = proportion of variance explained. r=−0.72 → r²=0.5184 → 51.84% explained
💡 Explanation
The coefficient of determination \(r^2\) (not \(r\)) tells us the proportion of variance in \(y\) explained by the linear model.

\(r^2 = (-0.72)^2 = 0.5184\) → the model explains approximately 51.8% of variation in \(y\).

Trap: \(r^2\) is always non-negative. The negative sign of \(r\) means negative linear relationship (as \(x\) increases, \(y\) decreases), but \(r^2\) is still positive.
§6 · Chi-Squared Test of Independence
Q15
Expected Frequencies & \(\chi^2\) Test Hard

A survey investigates whether gender (Male/Female) is associated with preferred genre (Action/Romance/Comedy). The observed contingency table has row totals 120 (Male) and 80 (Female), column totals 90 (Action), 60 (Romance), 50 (Comedy), and grand total 200. What is the expected frequency for Male–Action?

E = (row total × column total) / grand total — ALWAYS use this formula for expected frequencies
💡 Explanation
\(E_{\text{Male, Action}} = \dfrac{\text{Row total (Male)} \times \text{Column total (Action)}}{\text{Grand total}} = \dfrac{120 \times 90}{200} = \dfrac{10800}{200} = \mathbf{54}\)

Trap: Some students compute \(\dfrac{90}{200} \times 120 = 54\) correctly but then accidentally swap row/column totals when computing other cells.
Q16
Degrees of Freedom & Hypothesis Conclusion Tricky

A \(\chi^2\) test of independence is performed on a 3×4 contingency table at a 5% significance level. The calculated test statistic is \(\chi^2_{\text{calc}} = 14.3\). The critical value is \(\chi^2_{\text{crit}} = 12.592\). What is the conclusion of the test?

df = (rows−1)(cols−1) | Reject H₀ if χ²_calc > χ²_crit — "sufficient evidence of association"
💡 Explanation
\(df = (3-1)(4-1) = 2 \times 3 = 6\). \(\chi^2_{\text{crit}} = 12.592\) at 5% level with 6 df ✓

Since \(14.3 > 12.592\), we reject \(H_0\).

Correct wording: "There is sufficient evidence at the 5% significance level to suggest an association between the two variables."

Trap: Never say the test "proves" a causal link — \(\chi^2\) tests association only, not causation.
§7 · Spearman's Rank Correlation
Q17
Spearman vs Pearson — When to Use Which Concept Trap

A researcher ranks 8 students by their performance on a practical skill (ranks 1–8) and records their scores on a written test. The data are clearly non-linear. Which correlation coefficient is most appropriate, and what does a Spearman coefficient of \(r_s = -0.90\) indicate?

Spearman = rank-based → use for non-linear, ordinal, or outlier-heavy data. Negative rs = inverse monotonic trend
💡 Explanation
Spearman's \(r_s\) is appropriate when:
• Data are ordinal (ranks)
• Relationship is non-linear but monotonic
• Outliers may distort Pearson's \(r\)

\(r_s = -0.90\) indicates a strong negative monotonic relationship: as practical skill rank increases, written test score tends to decrease.

Key: Spearman measures monotonic (not necessarily linear) association.
Q18
Calculating Spearman's \(r_s\) with Tied Ranks Hard

Five students are ranked by two judges. Judge A's ranks: 1, 2, 3, 4, 5. Judge B's ranks: 2, 1, 4, 3, 5. Using the formula \(r_s = 1 - \dfrac{6\sum d_i^2}{n(n^2-1)}\), calculate Spearman's rank correlation coefficient.

d = rank A − rank B for each item; square each d; sum them up; plug into formula
💡 Explanation
Differences \(d_i\): (1−2), (2−1), (3−4), (4−3), (5−5) = −1, 1, −1, 1, 0
\(d_i^2\): 1, 1, 1, 1, 0 → \(\sum d_i^2 = 4\)

\(r_s = 1 - \dfrac{6 \times 4}{5(25-1)} = 1 - \dfrac{24}{120} = 1 - 0.2 = \mathbf{0.9}\)

Strong positive rank agreement between the two judges.
§8 · t-Test & Hypothesis Testing
Q19
One-Sample t-Test — Interpreting p-value Hard

A manufacturer claims their batteries last an average of 500 hours. A sample of 20 batteries has a mean life of 487 hours and standard deviation 28 hours. A one-tailed t-test is performed at a 5% significance level. The \(p\)-value is 0.034. Which conclusion is correct?

p-value < α → reject H₀ | p-value ≥ α → fail to reject H₀. Never "accept H₀"!
💡 Explanation
\(H_0: \mu = 500\), \(H_1: \mu < 500\) (one-tailed, lower)

Since \(p = 0.034 < \alpha = 0.05\), we reject \(H_0\).

Correct conclusion: "At the 5% significance level, there is sufficient evidence to suggest that the mean battery life is less than 500 hours."

Traps:
• Never "accept \(H_0\)" — only "fail to reject"
• Rejecting \(H_0\) does not mean the claim is definitely false — just insufficient evidence to support it at this level
Q20
Type I & Type II Errors — Hardest Concept Hard

A researcher sets the significance level at \(\alpha = 0.01\) instead of the usual \(\alpha = 0.05\). Which of the following is a consequence of this change?

α = P(Type I error) | Decreasing α → harder to reject H₀ → ↓ Type I, ↑ Type II error risk
💡 Explanation
Type I error (α): Rejecting a true \(H_0\) (false positive). By setting \(\alpha = 0.01\), you directly reduce \(P(\text{Type I})\).

Type II error (β): Failing to reject a false \(H_0\) (false negative). With a stricter threshold (smaller \(\alpha\)), it's harder to reject \(H_0\), so there's a greater chance of missing a true effect → \(\beta\) increases.

This is the fundamental trade-off in hypothesis testing. Power = \(1 - \beta\), so decreasing \(\alpha\) also reduces test power.

Quiz Complete! 🎓

Here's how you did on IB Math AI HL Statistics

0/20
0% correct