Further Statistics — A-Level Further Mathematics Study Guide

For: A-Level Further Mathematics candidates sitting Further Mathematics.

Covers: All core Further Statistics syllabus content for A-Level Further Mathematics, including continuous random variables, statistical inference, chi-squared tests, non-parametric tests, and probability generating functions.

You should already know: Strong A-Level Mathematics Pure Mathematics 1, 2, and 3 foundation.

A note on the practice questions: All worked questions in the "Practice Questions" section below are original problems written by us in the A-Level Further Mathematics style for educational use. They are not reproductions of past Cambridge International examination papers and may differ in wording, numerical values, or context. Use them to practise the technique; cross-check with official Cambridge mark schemes for grading conventions.

1. What Is Further Statistics?

Further Statistics is the advanced branch of A-Level Further Mathematics that extends your A-Level Mathematics statistics knowledge to solve complex, real-world probabilistic and inferential problems that standard basic methods cannot address. It counts for 25% of your total A-Level Further Mathematics marks in most exam sittings, balances theoretical probabilistic modelling and practical hypothesis testing, and is sometimes referred to as advanced or inferential statistics in broader learning resources. All questions on this topic require you to apply core formulas to novel contexts, rather than just rote recall, so understanding the logic behind each method is critical for scoring full marks.

2. Continuous random variables and PDFs

You already know discrete random variables (which take countable integer values) from A-Level Mathematics; continuous random variables (CRVs) take values across an uncountable range, such as time, height, or mass, and are described by a probability density function (PDF) $f(x)$. The PDF has two non-negotiable properties:

1. $f(x)\ge 0$ for all $x$ in the domain of the variable (you cannot have negative probability)
2. The total integral across the full domain is 1: ${\int }_{-\infty }^{\infty }f(x)dx=1$, since the sum of all possible outcomes is 100% probability.

For CRVs, the probability of the variable taking a single exact value is always 0, so you will always integrate across an interval to find probability: $P(a<X<b)={\int }_a^{b}f(x)dx$.

Worked example

A CRV $X$ has PDF $f(x)=\frac{3}{8}{x}^2$ for $0\le x\le 2$, 0 otherwise. Verify it is a valid PDF and find $P(1<X<1.5)$.

1. Validity check: ${\int }_0^2\frac{3}{8}{x}^{2}dx=\frac{3}{8}\ast \frac{x^3}{3}{|}_0^2=\frac{8}{8}=1$, and $f(x)$ is non-negative across its domain, so it is valid.
2. Probability calculation: ${\int }_1^{1.5}\frac{3}{8}{x}^{2}dx=\frac{3}{8}\ast [\frac{3.375}{3}-\frac{1}{3}]=\frac{19}{64}\approx 0.297$.

Exam tip: Examiners regularly ask for the median or mode of a CRV: the median $m$ satisfies ${\int }_{-\infty }^{m}f(x)dx=0.5$, while the mode is the value of $x$ that maximises $f(x)$ in the domain.

3. Inference — confidence intervals and tests for the mean

Statistical inference uses sample data to draw conclusions about an unknown population parameter, most commonly the population mean $\mu$ for A-Level Further Mathematics questions.

• Confidence intervals: If population variance ${\sigma }^2$ is known, use the standard normal Z-distribution: the 95% confidence interval is $\bar{x}\pm 1.96\ast \frac{\sigma }{\sqrt{n}}$, where $\bar{x}$ is the sample mean and $n$ the sample size. If population variance is unknown and the sample size is small (<30), use the t-distribution with $n-1$ degrees of freedom, so the interval becomes $\bar{x}\pm t_{n-1,\frac{\alpha }{2}}\ast \frac{s}{\sqrt{n}}$, where $s$ is the sample standard deviation.
• Hypothesis tests for the mean: Set a null hypothesis $H_0:\mu ={\mu }_0$ and one or two-tailed alternative hypothesis, calculate the test statistic (Z if ${\sigma }^2$ is known, $t=\frac{\bar{x}-{\mu }_0}{s/\sqrt{n}}$ if ${\sigma }^2$ is unknown), and compare to the critical value for your significance level to decide whether to reject $H_0$.

Worked example

A sample of 10 students has an average test score of 72, with sample standard deviation 6. Assuming scores are normally distributed, find the 95% confidence interval for the population mean.

• Unknown population variance, small sample: use t-distribution with 9 degrees of freedom, $t_{9,0.025}=2.262$.
• Interval = $72\pm 2.262\ast \frac{6}{\sqrt{10}}=72\pm 4.30$, so the final interval is $(67.7,76.3)$.

4. Chi-squared tests — goodness-of-fit and contingency

Chi-squared (${\chi }^2$) tests are non-parametric tests used to compare observed frequency data to expected values under a null hypothesis. There are two core variants for A-Level Further Mathematics:

1. Goodness-of-fit test: Tests if observed data fits a hypothesised distribution (e.g. uniform, binomial, normal). The test statistic is ${\chi }^2=\sum \frac{(O_i-E_i{)}^2}{E_i}$, where $O_i$ is observed frequency and $E_i$ expected frequency for each category. Degrees of freedom $\nu =k-1-m$, where $k$ is the number of categories and $m$ the number of parameters estimated from sample data.
2. Contingency table test: Tests if two categorical variables are independent. For an $r\times c$ contingency table, degrees of freedom are $\nu =(r-1)(c-1)$.

For both tests, all expected frequencies must be at least 5: if any $E_i<5$, combine adjacent categories before calculating the test statistic, and adjust your degrees of freedom accordingly.

Worked example

You roll a die 60 times, with observed counts: 1:8, 2:12, 3:11, 4:9, 5:10, 6:10. Test at 5% significance if the die is fair.

• $H_0$: die is fair, so each outcome has probability $\frac{1}{6}$, expected count per category = 10.
• ${\chi }^2=\frac{(8-10{)}^2}{10}+\frac{(12-10{)}^2}{10}+\frac{(11-10{)}^2}{10}+\frac{(9-10{)}^2}{10}+0+0=1.0$
• Degrees of freedom $\nu =6-1=5$, critical value at 5% = 11.07. Since 1.0 < 11.07, you do not reject $H_0$: the die is fair.

5. Non-parametric tests

Non-parametric tests do not assume your underlying data follows a normal distribution, unlike Z or t-tests, so they are used for small or skewed datasets. The two core tests for A-Level Further Mathematics are:

1. Wilcoxon signed-rank test: Used for paired data or tests of a single sample median. Calculate differences between paired observations, ignore zero differences, rank absolute differences from smallest to largest (average ranks for tied values), assign the sign of the original difference to each rank, sum positive ranks $T_{+}$ and negative ranks $T_{-}$. The test statistic is the smaller of $T_{+}$ and $T_{-}$, and you reject $H_0$ if the test statistic is less than the critical value from the Wilcoxon table.
2. Wilcoxon rank-sum test: Used for two independent samples. Combine the two samples, rank all values from smallest to largest, sum the ranks for the smaller sample to get your test statistic $W$, and reject $H_0$ if $W$ falls outside the critical range for your sample sizes.

Worked example

8 students take a test before and after a revision course, with differences (after - before) = [3, 2, -1, 4, 5, 2, 1, 3]. Test at 5% one-tailed significance if the course improves scores.

• $H_0$: median difference = 0, $H_1$: median difference > 0.
• Rank absolute differences, average ties: the two differences of 1 get ranks 1.5, the two differences of 2 get ranks 3.5, the two differences of 3 get ranks 5.5, with 4 and 5 getting ranks 7 and 8 respectively.
• Sum of negative ranks $T_{-}=1.5$, test statistic = 1.5. The one-tailed critical value for $n=8$ at 5% is 6. Since 1.5 < 6, reject $H_0$: the course improves scores.

6. Probability generating functions

Probability generating functions (PGFs) are a algebraic tool to find the distribution, mean, and variance of discrete random variables that take non-negative integer values. For a discrete random variable $X$, the PGF is defined as $G_X(t)=E(t^X)={\sum }_{x=0}^{\infty }P(X=x)t^x$. Key properties you will be tested on:

1. $G_X(1)=1$, as the sum of all probabilities is 1.
2. Expected value: $E(X)=G_X^{\prime }(1)$, the first derivative of the PGF evaluated at $t=1$.
3. Variance: $Var(X)=G_X^{\prime \prime }(1)+G_X^{\prime }(1)-[G_X^{\prime }(1){]}^2$, using the first and second derivatives evaluated at $t=1$.
4. For independent random variables $X$ and $Y$, the PGF of their sum is $G_{X+Y}(t)=G_X(t)G_Y(t)$.

Worked example

$X\sim Binomial(n=3,p=0.5)$. Find its PGF, then use it to calculate $E(X)$ and $Var(X)$.

• $P(X=x)=\left(\genfrac{}{}{0ex}{}{3}{x}\right)0.5^3$, so $G_X(t)={\sum }_{x=0}^3\left(\genfrac{}{}{0ex}{}{3}{x}\right)0.5^3{t}^x=0.5^3(1+t{)}^3=\frac{(1+t{)}^3}{8}$.
• First derivative $G_X^{\prime }(t)=\frac{3(1+t{)}^2}{8}$, so $G_X^{\prime }(1)=\frac{12}{8}=1.5=E(X)$, matching the binomial formula $np$.
• Second derivative $G_X^{\prime \prime }(t)=\frac{6(1+t)}{8}$, so $G_X^{\prime \prime }(1)=1.5$. $Var(X)=1.5+1.5-(1.5{)}^2=0.75$, matching the binomial formula $np(1-p)$.

7. Common Pitfalls (and how to avoid them)

• Wrong move: Using the Z-distribution for small sample confidence intervals when population variance is unknown. Why: Students confuse Z and t use cases. Correct move: Use the t-distribution with $n-1$ degrees of freedom if variance is unknown and $n<30$, and explicitly state the distribution you are using to earn the method mark.
• Wrong move: Forgetting to combine categories in chi-squared tests when expected frequencies are less than 5. Why: Students rush through expected value calculations. Correct move: Check every $E_i$ before calculating the test statistic, combine adjacent small categories, and reduce your degrees of freedom by 1 for every category you combine.
• Wrong move: Assigning sequential ranks to tied values in Wilcoxon tests. Why: Students forget the rank averaging rule. Correct move: For tied values, assign the average of the ranks they would have taken, and retain the sign of the difference for signed-rank tests.
• Wrong move: Calculating $P(X=a)$ for a CRV by plugging $a$ into the PDF. Why: Students mix up PDFs for continuous variables and PMFs for discrete variables. Correct move: Remember single-point probability for CRVs is always 0, so you must integrate across an interval to find probability.
• Wrong move: Omitting the $[G^{\prime }(1){]}^2$ term when calculating variance from a PGF. Why: Students misremember the PGF variance formula. Correct move: Write the full variance formula at the start of every PGF question to avoid omission.

8. Practice Questions (A-Level Further Mathematics Style)

Question 1

A continuous random variable X has PDF $f(x)=kx(4-x)$ for $0\le x\le 4$, 0 otherwise. (a) Find the value of $k$. (b) Calculate $P(X>1)$. (c) Find the median of $X$.

Solution

(a) Total integral over domain = 1: $${\int }_0^{4}kx(4-x)dx=k[2x^2-\frac{x^3}{3}{]}_0^4=k\ast \frac{32}{3}=1⟹k=\frac{3}{32}$$ (b) Integrate from 1 to 4: $$P(X>1)=\frac{3}{32}[2x^2-\frac{x^3}{3}{]}_1^4=\frac{3}{32}\ast \frac{27}{3}=\frac{27}{32}=0.84375$$ (c) Solve ${\int }_0^{m}f(x)dx=0.5$: $$\frac{3}{32}[2m^2-\frac{m^3}{3}]=0.5⟹m^3-6m^2+16=0$$ The only solution between 0 and 4 is $m=2$, so the median is 2.

Question 2

A 2x3 contingency table has observed values: [12, 18, 10] for row 1, [8, 12, 20] for row 2. Test at 5% significance if the two variables are independent.

Solution

$H_0$: variables are independent, $H_1$: variables are dependent. Row totals = 40, 40; column totals = 20, 30, 30; grand total = 80. Expected frequencies: [10, 15, 15] for row 1, [10,15,15] for row 2. $${\chi }^2=\frac{(12-10{)}^2}{10}+\frac{(18-15{)}^2}{15}+\frac{(10-15{)}^2}{15}+\frac{(8-10{)}^2}{10}+\frac{(12-15{)}^2}{15}+\frac{(20-15{)}^2}{15}=5.33$$ Degrees of freedom = $(2-1)(3-1)=2$, critical value at 5% = 5.991. Since 5.33 < 5.991, do not reject $H_0$: variables are independent.

Question 3

The PGF of a discrete random variable X is $G_X(t)=\frac{t}{(2-t{)}^2}$. Find (a) $E(X)$, (b) $Var(X)$.

Solution

(a) First derivative using quotient rule: $$G_X^{\prime }(t)=\frac{(2-t{)}^2+2t(2-t)}{(2-t{)}^4}=\frac{2+t}{(2-t{)}^3}$$ Evaluate at $t=1$: $E(X)=\frac{3}{1}=3$. (b) Second derivative: $$G_X^{\prime \prime }(t)=\frac{(2-t{)}^3+3(2+t)(2-t{)}^2}{(2-t{)}^6}=\frac{(2-t)+3(2+t)}{(2-t{)}^4}$$ Evaluate at $t=1$: $G_X^{\prime \prime }(1)=1+9=10$. $$Var(X)=10+3-(3{)}^2=4$$

9. Quick Reference Cheatsheet

10. What's Next

The content in this guide forms the foundation for all advanced statistics questions in your A-Level Further Mathematics exam, and also connects directly to optional further topics including linear regression, correlation, and stochastic processes if you choose to pursue higher level mathematics or statistics at university. Mastering these concepts will also make it easier to tackle cross-topic questions that combine statistics with pure mathematics content like calculus or series, which are common in the 8-10 mark higher-order questions on the paper.

If you struggle with any of the subtopics covered here, or want to practice more exam-style questions tailored to your weak areas, you can ask Ollie for personalized help, additional practice problems, or step-by-step explanations of any concept at any time. Head to the homepage to access all free A-Level Further Mathematics study resources, including past paper solutions and timed mock exams.

Aligned with the Cambridge International AS & A Level Further Mathematics 9231 syllabus. OwlsAi is not affiliated with Cambridge Assessment International Education.