MATH2089 Numerical Methods & Statistics
The Complete Exam Bible · T3 2026MATH2089 Overview
- UNSW Sydney
- Term 3 2026
- Level 2 course
- Numerical and statistical computing
- In person lab assessments
MATH2089 Numerical Methods & Statistics combines two technical toolkits that engineers use to turn mathematical models and data into defensible decisions.
- Assessed by Online work, lab activity, tests and two final components
- Hardest step Choosing a valid method before calculating
- How to prepare Recompute examples and apply an independent check
What you get in the full MATH2089 Exam Bible
- 14 chapters, 57 pages — every examinable topic
- 36 original practice questions, each with a model answer in the Final Exam format
- 14 worked examples, solved the way MATH2089 marks
- 14 computed figures, redrawn from the course’s own numbers
- Exam-morning plan for the Final Exam (50%) — what to revise last, where marks are lost
Free on this page: assessment table, chapter map, one worked example, key terms, FAQ. The rest is in the full exam bible.
How MATH2089 is assessed
| Component | Weight | Format |
|---|---|---|
| Online Quizzes · hurdle | 20% | Individual online activities across the Numerical Methods and Statistics components |
| Active Learning Lab Tutorials · hurdle | 10% | In person lab activity across both course components |
| Lab Tests · hurdle | 20% | Numerical Methods tests and the Statistics midterm in computing labs on Mobius |
| Final Exam · hurdle | 50% | Two in person one hour Mobius components in computing labs |
To pass MATH2089, you must achieve at least 50% in the final combined course mark and at least 40%, or 20 out of 50, in each component. This per-component minimum applies separately to the Numerical Methods half and the Statistics half; every grouped assessment row contributes to both half marks.
What other students are asking about MATH2089
- Where can I find lecture notes or study guides for MATH2089 at UNSW?
- The MATH2089 study guide in AskSia Library covers Numerical Methods & Statistics in 14 chapters, with the assessment table, a chapter-by-chapter topic map, key terms and a worked example, free to read on this page.
- Is there a final exam in MATH2089, and how much is it worth?
- Yes. In the offering this AskSia Library guide is built from, Final Exam carries 50% of the MATH2089 mark; the full split is Online Quizzes 20%, Active Learning Lab Tutorials 10%, Lab Tests 20%, Final Exam 50%.
- What does MATH2089 cover on linear systems and conditioning?
- In AskSia Library's MATH2089 guide, the Linear Systems and Conditioning chapter explains sensitivity, LU and Cholesky factorisation, sparse structure and residual checks.
- What does MATH2089 cover on least squares and interpolation?
- In AskSia Library's MATH2089 guide, the Least Squares and Interpolation chapter explains orthogonal projection, QR stability, Lagrange polynomials and Horner evaluation.
- What does MATH2089 cover on nonlinear equations?
- In AskSia Library's MATH2089 guide, the Nonlinear Equations chapter explains bisection, fixed point, Newton and secant convergence with stopping checks.
Current MATH2089 dates
| Date | Item | Control |
|---|---|---|
| Week 5 | Numerical Methods Test 1 | Held during the scheduled tutorial. |
| Week 7 | Statistics Midterm Test | Held during the scheduled Statistics tutorial. |
| Week 9 | Numerical Methods Test 2 | Held during the scheduled tutorial. |
| Term 3 examination period | Final Exam | Two one hour components; confirm the official timetable. |
Dates are as published in the Term 3 2026 MATH2089 Moodle assessment tables. Confirm exact deadlines and submission settings in the live LMS.
What MATH2089 covers
MATH2089 gives equal weight to Numerical Methods and Statistics. The first half develops error analysis, linear algebra, approximation, differential equations and PDE discretisation. The second half moves from probability and distributions through sampling, inference, regression and ANOVA. The guide connects both halves through method conditions, visible calculations and independent checks.
Numerical Computing and Error
Floating point representation, error scales, stability and computational costFloating point arithmetic replaces the real number line with a finite representable set, so every numerical result must be read through an error model.
- Floating point spacing and finite precision
- Signed, absolute and relative error
- Roundoff, truncation and cancellation
- Floating point operation counts
- Big O growth and resource tradeoffs
- Stability and independent checks
Linear Systems and Conditioning
Sensitivity, LU and Cholesky factorisation, sparse structure and residual checksLinear systems combine a mathematical sensitivity question with an algorithmic solve. The condition number estimates how data perturbations can affect a solution, while LU and Cholesky factorisations organise the computation into triangular solves.
- Vector and matrix norms
- Condition numbers and sensitivity
- LU factorisation and pivoting
- Forward and back substitution
- Cholesky eligibility and reuse
- Sparse and banded matrix structure
Least Squares and Interpolation
Orthogonal projection, QR stability, Lagrange polynomials and Horner evaluationInterpolation and least squares use polynomial language for different objectives. Interpolation matches every exact node, while least squares accepts residuals and minimises their squared length for an overdetermined system.
- Overdetermined systems and residuals
- Orthogonal projection
- Normal equations
- QR factorisation and conditioning
- Lagrange interpolation
- Horner polynomial evaluation
Nonlinear Equations
Bisection, fixed point, Newton and secant convergence with stopping checksRoot finding balances global guarantees against local speed.
- Sign changing brackets
- Bisection error bounds
- Fixed point contraction
- Newton iteration
- Secant updates
- Convergence order and stopping
Differentiation and Quadrature
Taylor error, finite differences and weighted integration rulesTaylor expansion derives finite difference and quadrature formulas and exposes their leading truncation errors. Numerical differentiation must balance decreasing truncation against increasing roundoff at very small steps.
- Taylor polynomial and remainder
- Forward and centred differences
- Roundoff versus truncation
- Composite trapezoidal rule
- Simpson's rule
- Gauss Legendre nodes and weights
Ordinary Differential Equations
Euler, Heun, Runge Kutta, stability, stiffness and shootingInitial value problems become discrete time sequences through numerical updates. Euler's method uses the current slope and has first order global accuracy, while Heun and Runge Kutta methods combine staged slopes for higher order.
- Initial value problems
- Euler updates
- Heun and Runge Kutta stages
- Global error and refinement
- Stiffness and implicit methods
- Shooting for boundary values
Partial Differential Equations
Elliptic, parabolic and hyperbolic classification with finite differencesSecond order PDE classification separates elliptic, parabolic and hyperbolic behaviour through the principal part. The class determines which boundary or initial data are natural and how a finite difference method should be assembled.
- PDE discriminant classification
- Elliptic boundary problems
- Parabolic diffusion problems
- Hyperbolic wave problems
- Five point finite difference stencil
- Sparse assembly and stability
Probability and Descriptive Statistics
Sampling design, event rules, graphical summaries and numerical descriptionStatistical analysis begins with how observations were produced. This chapter distinguishes population from sample and parameter from statistic, then connects random sampling with the scope of inference.
- Populations, samples and parameters
- Random sampling and bias
- Mean, median and quantiles
- Variance and standard deviation
- Histograms, boxplots and scatterplots
- Conditional probability and independence
Random Variables and Distributions
Discrete and continuous models, expectation, variance, probabilities and quantilesRandom variables connect uncertain outcomes with numerical models. Discrete variables use probability masses and continuous variables use densities whose interval areas are probabilities.
- Discrete and continuous variables
- Mass, density and cumulative functions
- Expectation and variance
- Binomial and Poisson models
- Uniform and exponential models
- Normal probabilities and quantiles
Sampling and Confidence Intervals
Sampling distributions, standard errors, coverage and sample size planningSampling distributions describe how statistics vary across repeated samples and provide the standard errors used in inference. The central limit theorem supports approximate normal behaviour for sample means under suitable conditions.
- Sampling distributions
- Central limit theorem
- Standard error
- Known variance z intervals
- Unknown variance t intervals
- Proportion intervals and sample size
Hypothesis Testing
Null models, alternatives, p values, error types and powerHypothesis tests compare observed evidence with a null model through a standardised statistic and a preselected alternative.
- Null and alternative hypotheses
- One sided and two sided tests
- Test statistics
- P values and significance
- Type I and Type II errors
- Power and confidence interval links
Comparing Means
Paired and independent designs, standard errors and difference intervalsComparisons of two means begin with the study design. Repeated or matched units produce paired differences and reduce the problem to one sample of changes.
- Independent sample designs
- Paired and matched designs
- Difference direction
- Welch two sample method
- Pooled variance assumption
- Confidence intervals for differences
Linear Regression
Least squares, coefficient inference, residual diagnostics and predictionSimple linear regression models a mean response as a line plus random error. Least squares estimates intercept and slope by minimising squared residuals, while inference uses standard errors and a t distribution under the error assumptions.
- Simple linear model
- Least squares coefficients
- Fitted values and residuals
- Slope inference
- Residual diagnostics
- Confidence and prediction intervals
Analysis of Variance
Between and within variation, the F ratio and adjusted follow upOne way ANOVA compares several population means through a single decomposition of variability. Between group variation reflects separation among sample means, while within group variation reflects residual noise.
- Equal means null model
- Between group variation
- Within group variation
- Mean squares and F ratio
- ANOVA assumptions
- Adjusted follow up comparisons
The Numerical Methods half starts with floating point representation, error and computational cost, then develops linear systems, conditioning, least squares, interpolation, nonlinear equations, numerical differentiation, quadrature, ordinary differential equations and finite difference methods for partial differential equations.
The Statistics half begins with probability, sampling and descriptive summaries, then studies random variables and named distributions, sampling distributions, confidence intervals, hypothesis tests, comparisons of means, linear regression and analysis of variance. Across both halves, calculation alone is not enough.
A numerical algorithm needs an eligibility condition, an error or stability argument and a check such as a residual or refinement ratio. A statistical procedure needs a design, a sampling model, a measure of uncertainty and a conclusion whose scope matches the evidence.
The official UNSW Handbook groups the assessment into Online Quizzes worth 20%, Active Learning Lab Tutorials worth 10%, Lab Tests worth 20% and a Final Exam worth 50%. Moodle shows how those grouped rows split evenly across Numerical Methods and Statistics. The final consists of separate one hour components in computing labs on Mobius.
Numerical Methods covers Topics 1 to 7 with heavy emphasis on Topic 7, while Statistics covers all its material with heavy emphasis on statistical inference. The course also has a component requirement: the combined mark must reach the course pass level and each half must reach its published minimum.
This guide therefore treats the two halves as equally important and uses original questions rather than reproducing course assessment stems.
Worked example · free
Bisection: preserve a bracket and certify the error
- +1The starting bracket width is 2 minus 1, which equals 1.
- +1Bisection halves the bracket on every step while preserving a sign change.
- +1After three steps, the width is 1 divided by 2 cubed, which equals 0.125.
- +1The root remains inside that final bracket, giving a direct location certificate independent of derivative values.
Key terms
- Condition number
- A sensitivity measure that bounds how input perturbations may be amplified in a numerical solution.
- Residual
- The difference between an observed or required value and the value produced by a fitted model or approximate solution.
- Roundoff error
- Error introduced because a computer stores and operates on finitely represented numbers.
- Truncation error
- Error created when an infinite or exact mathematical process is replaced by a finite approximation.
- Convergence order
- The asymptotic rate at which approximation error decreases as iterations or discretisation are refined.
- Sampling distribution
- The probability distribution of a statistic across repeated samples drawn under the same design.
- Standard error
- The standard deviation of a statistic's sampling distribution, used to quantify sampling uncertainty.
- Confidence interval
- A range produced by a repeated sampling procedure to estimate plausible values of a population parameter.
- P value
- The probability under a null model of a test statistic at least as extreme as the observed result.
- Statistical power
- The probability that a test rejects the null when a specified alternative is true.
- Paired design
- A comparison design in which observations are linked within the same or deliberately matched units.
- Least squares
- A fitting criterion that chooses parameters to minimise the sum of squared residuals.
- Prediction interval
- An interval for one future observation that includes both mean estimation uncertainty and individual variation.
- F statistic
- A ratio of model or between group variation to residual or within group variation used in ANOVA.
MATH2089 FAQ
How is this course assessed?
The UNSW Handbook groups the course into Online Quizzes worth 20%, Active Learning Lab Tutorials worth 10%, Lab Tests worth 20% and a Final Exam worth 50%. Moodle provides the detailed split across the equal Numerical Methods and Statistics components, and the grouped weights total 100%.
What is the final exam format?
The Term 3 course materials publish two final components on Mobius in computing labs. Numerical Methods lasts one hour and covers Topics 1 to 7 with heavy emphasis on Topic 7. Statistics lasts one hour and covers all material with heavy emphasis on statistical inference.
Is there a hurdle requirement?
Yes. The Moodle assessment hub states that the combined course mark must reach at least 50%, and each equal component must reach at least 20 out of 50. This is a course component requirement, not a hurdle attached only to one exam row.
Why do students find the course demanding?
The workload spans numerical algorithms and statistical inference, and both halves require method selection before arithmetic. Marks can be lost through a stability condition, a wrong probability boundary, an ignored sampling design or an unsupported interpretation even when calculator steps look plausible.
Can I use MATLAB or Python?
The course materials support both MATLAB and Python, and the source states that software installed on the lab computers is available during assessments. Practise the commands you need, but keep the mathematical setup visible so software output can be checked and explained.
Are past final papers included?
No released historical final paper appears in the available course materials. They do include Numerical Methods revision tests, practice tests, a Statistics midterm practice activity and worked practice class solutions. This guide uses their question types to create new questions without copying assessment stems.
Questions students ask in each MATH2089 chapter
- Why is the smallest step not always best?
- A smaller step can reduce truncation error, but it can also magnify roundoff through subtraction and division. Numerical accuracy often improves only until an optimal region, after which finite precision begins to dominate. Chapter 01 · Numerical Computing and Error
- Does a tiny residual prove the solution is accurate?
- Not by itself. An ill conditioned matrix can permit a small residual while the solution changes substantially. Interpret the residual together with the condition number and the scale of input uncertainty. Chapter 02 · Linear Systems and Conditioning
- Why is QR preferred to normal equations?
- Forming the normal equations squares the two norm condition number of the design matrix. QR solves an equivalent projection problem without that extra sensitivity amplification. Chapter 03 · Least Squares and Interpolation
- Why can Newton's method fail?
- A poor start can enter another attraction basin, a small derivative can create a huge step, and difficult geometry can cause divergence or cycling. Local speed does not give a global guarantee. Chapter 04 · Nonlinear Equations
- Why can differentiation worsen with smaller h?
- A smaller h reduces truncation initially, but nearby function values become almost equal. Their subtraction loses significant digits and division by h magnifies the roundoff. Chapter 05 · Differentiation and Quadrature
- What does first order global error mean?
- For a smooth stable problem, Euler error at a fixed final time is proportional to h. Halving h should therefore roughly halve the leading error in the asymptotic regime. Chapter 06 · Ordinary Differential Equations
- What does the PDE discriminant decide?
- The sign of B squared minus 4AC classifies the principal part as elliptic, parabolic or hyperbolic. That class signals different data patterns and numerical behaviour. Chapter 07 · Partial Differential Equations
- Can a large sample remove selection bias?
- No. A large sample reduces random variability within the sampling frame, but it does not repair systematic exclusion, nonresponse or measurement bias. Design determines representativeness. Chapter 08 · Probability and Descriptive Statistics
- Why is a continuous point probability zero?
- Probability for a continuous variable is area under a density over an interval. A single point has zero width, even when the density value there is high. Chapter 09 · Random Variables and Distributions
- What does 95% confidence mean?
- It describes the long run coverage of the interval procedure across repeated samples. It is not a 95% probability statement about a fixed parameter after one interval is observed. Chapter 10 · Sampling and Confidence Intervals
- Is a p value the probability the null is true?
- No. It is calculated assuming the null model and measures how extreme the observed statistic is under that assumption. It does not assign probability to the hypothesis. Chapter 11 · Hypothesis Testing
- How do I identify paired data?
- Look for repeated measurements on the same units or deliberate matching across units. The analysis should use one signed difference per pair rather than treat all observations as independent. Chapter 12 · Comparing Means
- What does R squared prove?
- It measures the proportion of sample response variation explained by the fitted linear relation. It does not prove causation, validate assumptions or guarantee accurate extrapolation. Chapter 13 · Linear Regression
- What does ANOVA rejection establish?
- It establishes evidence that not all population means are equal. It does not identify the differing groups and does not imply that every pair differs. Chapter 14 · Analysis of Variance
How to study for the exam
Treat the course as two equal revision programmes connected by one habit: every method needs an independent check. For Numerical Methods, build a table with the method, eligibility conditions, dominant error, cost and verification. Recompute small examples for LU or Cholesky solves, QR least squares, bisection and Newton updates, quadrature refinement, Euler and Runge Kutta steps, and PDE stencils.
Give extra time to Topic 7 because the published Numerical Methods final emphasises it. For Statistics, begin every question with the observational unit, design, parameter and model. Practise event translation for distributions, then connect sampling distributions to confidence intervals and tests.
Keep paired and independent comparisons separate, diagnose residuals before regression inference, and read ANOVA as a global test followed by adjusted comparisons. Rehearse in the lab software you intend to use, because the tests and final components run on Mobius in computing labs and handheld calculators are not permitted.
The final week should alternate one Numerical Methods method set with one Statistics inference set so neither component falls below its published minimum.
Your AI Mathematics tutor for MATH2089
Stuck on a hard MATH2089 question? Sia is AskSia’s AI Mathematics tutor — ask any MATH2089 Numerical Methods & Statistics question and get a clear, step-by-step explanation grounded in how the course is actually taught and assessed. Read this whole study guide free, then take your hardest questions to Sia.