PHAS0083 Statistical Analysis of Data
PHAS0083 Overview
- 2026/27 academic year
- University College London
- Graduate-level physics module
- In-person written final
PHAS0083 Statistical Analysis of Data is a graduate-level UCL module that sets out the mathematical framework of statistics and the techniques most used in physics: fitting models, testing hypotheses and estimating uncertainties. The lecture notes describe two directions of reasoning.
- Assessed by Five coursework quizzes worth 50% in total and an in-person written final exam worth 50%.
- Core idea Run models forwards to predict data, then reason backwards from data to parameters.
- Hardest step Turning a physical situation into a model with a named parameter before computing.
- Linking tool The likelihood connects estimation, likelihood ratio tests and confidence intervals.
What you get in the full PHAS0083 Exam Bible
- 9 chapters, 38 pages — every examinable topic, week by week
- 34 original practice questions, most with model answers in the Final Exam format
- 9 worked examples, solved the way PHAS0083 marks
- 37 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 PHAS0083 is assessed
| Component | Weight | Format |
|---|---|---|
| Coursework 1 | 5% | Online coursework quiz closing Monday 19 October 2026. One attempt; algebra is entered in STACK format. |
| Coursework 2 | 5% | Online coursework quiz closing Monday 2 November 2026. One attempt. |
| Coursework 3 | 10% | Online coursework quiz closing Monday 16 November 2026, with a supplied penguin data file. |
| Coursework 4 | 15% | Online coursework quiz due Monday 30 November 2026. |
| Coursework 5 | 15% | Online coursework quiz due Monday 14 December 2026. |
| Final Exam | 50% | In person and written. Anything covered in the notes or lectures can be examined; long derivations are not expected. Date set by the UCL examination timetable. |
The assessment page gives weekly assessment 50% and the exam 50%, and its table lists the five coursework weights. The hurdle status of each component is not stated in the module information for this offering; check the module page on the course site for any minimum-mark condition.
Assessment structure
What other students are asking about PHAS0083
- Where can I find lecture notes or study guides for PHAS0083 at ucl?
- The PHAS0083 study guide in AskSia Library covers Statistical Analysis of Data in 9 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 PHAS0083, and how much is it worth?
- Yes. In the offering this AskSia Library guide is built from, Final Exam carries 50% of the PHAS0083 mark; the full split is Coursework 1 5%, Coursework 2 5%, Coursework 3 10%, Coursework 4 15%, Coursework 5 15%, Final Exam 50%.
- What does PHAS0083 cover on common families of distributions?
- In AskSia Library's PHAS0083 guide, the Common Families of Distributions chapter explains week 2. Expectation and variance, transformations, the uniform, binomial, Poisson, exponential and normal families and how they connect.
- What does PHAS0083 cover on monte carlo simulation and density estimates?
- In AskSia Library's PHAS0083 guide, the Monte Carlo Simulation and Density Estimates chapter explains week 2. Inverse-cdf and accept-reject sampling, hierarchical simulation, the empirical cdf, histograms and kernel density estimates.
- What does PHAS0083 cover on joint, marginal and conditional distributions?
- In AskSia Library's PHAS0083 guide, the Joint, Marginal and Conditional Distributions chapter explains week 3. Random vectors, joint pmf and pdf, marginalisation, conditional densities, independence and change of variables.
Current PHAS0083 dates
| Date | Item | Control |
|---|---|---|
| 19 October 2026 | Coursework 1 closes | The quiz opened on 13 October and closed at 5:00 pm. |
| 2 November 2026 | Coursework 2 closes | The quiz opened on 26 October and closed at 5:00 pm. |
| 16 November 2026 | Coursework 3 closes | The quiz opened on 9 November and closed at 5:00 pm. |
| 30 November 2026 | Coursework 4 deadline | Weight 15% of the final grade. |
| 14 December 2026 | Coursework 5 deadline | Weight 15% of the final grade. |
Dates are as published in the 2026/27 module assessment page. Confirm exact deadlines and submission settings in the live LMS.
What PHAS0083 covers
Nine chapters follow the lecture notes: five build models that generate data, and four reverse the direction to estimate, test and bound parameters from an observed sample.
Statistical Inference and Probability Theory
Week 1. Forward and backward flow, the four tasks of inference, sample spaces, conditional probability, Bayes' rule and random variables.Statistical Inference and Probability Theory sets up the language for the rest of the module.
- Two sources of randomness
- Deductive and inductive reasoning
- The four tasks of statistical inference
- Sample spaces, events and the axioms
- Conditional probability and Bayes' rule
- Independence and counting rules
Common Families of Distributions
Week 2. Expectation and variance, transformations, the uniform, binomial, Poisson, exponential and normal families and how they connect.Common Families of Distributions gives the building blocks of physics models.
- Expectation, variance and moments
- Functions of a random variable
- Uniform and Bernoulli distributions
- Binomial counts and their mode
- Poisson counts and the binomial limit
- Exponential waiting times
Monte Carlo Simulation and Density Estimates
Week 2. Inverse-cdf and accept-reject sampling, hierarchical simulation, the empirical cdf, histograms and kernel density estimates.Monte Carlo Simulation and Density Estimates brings computation in early so that later theory can be checked against draws.
- Pseudo-random numbers and seeds
- Probability integral transform
- Accept-reject sampling and efficiency
- Simulating hierarchical models
- The empirical cdf and quantiles
- Number of draws for a target precision
Joint, Marginal and Conditional Distributions
Week 3. Random vectors, joint pmf and pdf, marginalisation, conditional densities, independence and change of variables.Joint, Marginal and Conditional Distributions extends single-variable probability to random vectors.
- Random vectors and joint pmf or pdf
- Marginal distributions as projections
- Conditional distributions as slices
- Independence and the product rule
- Joint distributions of hierarchical models
- Change of variables with a Jacobian
Covariance, Correlation and Error Propagation
Week 3. Covariance and correlation, the variance of sums, and first-order propagation of uncertainty with correlated inputs.Covariance, Correlation and Error Propagation measures how strongly variables depend on each other and uses that measure to carry uncertainties through calculations.
- Covariance and its shortcut formula
- Correlation and linear association
- Uncorrelated versus independent
- The bivariate normal
- Variance of linear combinations
- Delta method with a covariance matrix
Random Samples and Convergence
Week 4. Population and sample, the sample mean and variance, the law of large numbers, the central limit theorem and simulated samples.Random Samples and Convergence turns the module from the forward flow to the backward flow.
- Random sample and population
- Statistics and their distributions
- Sample mean and its variance
- Sample variance and the n minus 1 divisor
- Sample correlation
- Law of large numbers
Point Estimation and Maximum Likelihood
Week 4. Estimators, bias, variance and mean squared error, the likelihood function, maximum likelihood and its large-sample behaviour.Point Estimation and Maximum Likelihood asks how to choose a single value for a parameter and how to judge the choice.
- Point estimators as statistics
- Bias, variance and mean squared error
- The likelihood function
- Maximum likelihood estimation
- Numerical minimisation of the negative log-likelihood
- Sufficiency and the likelihood principle
Hypothesis Testing and Likelihood Ratio Tests
Week 5. Null and alternative hypotheses, rejection regions, size, power, likelihood ratio tests and calibrating a test by simulation.Hypothesis Testing and Likelihood Ratio Tests builds the framework for deciding between hypotheses about a population parameter.
- Null and alternative hypotheses
- Rejection regions and test statistics
- Type I and Type II errors
- Power function, size and level
- Likelihood ratio tests and nuisance parameters
- Neyman-Pearson lemma and most powerful tests
Interval Estimation and Coverage
Week 6. Interval estimators, coverage probability, the confidence coefficient, pivots and confidence sets from inverted tests.Interval Estimation and Coverage replaces a single estimate with a set that comes with a frequency guarantee.
- Interval estimators and estimates
- Coverage probability
- Confidence coefficient
- One-sided limits
- Inverting tests and the confidence belt
- Numerical inversion on a grid
The forward direction runs from a model to the data sets it tends to produce, and is built from probability, the common families of distributions, Monte Carlo simulation and multivariate models. The backward direction runs from observed data to the model, and is what statistical inference means here: random samples, point estimation, hypothesis testing and interval estimation.
Assessment is split evenly.
Five coursework quizzes carry 50% of the final grade, weighted 5, 5, 10, 15 and 15 so that the later topics count for more, and an in-person written final exam carries the other 50%. The opening lecture states that exam questions will closely resemble the coursework, that anything covered in the notes or lectures can be examined, and that long derivations are not expected.
Coursework is answered online in a STACK question format, so algebra must be entered with the notation used in the question.
The practical habit the module rewards is to write the model first. Name the random variable and its support, choose the family and its parameter, state the target quantity, and only then compute.
Physics examples such as photon counts, decay times and repeated measurements recur throughout the notes, and many results are checked by simulating data from the model. This guide follows that sequence chapter by chapter, with original worked examples and practice.
Worked example · free
Write a Poisson model for detector counts and find a tail probability
- 2Name the random variable: X is the number of counts in one minute, with support 0, 1, 2 and so on. Independent events at a constant rate give X distributed as Poisson with mean 2.5. This allocation is original practice weighting, not a UCL marking scheme.
- 1Use the complement: P(X at least 5) = 1 minus P(X at most 4), with P(X = k) = exp(-2.5) 2.5ᵏ / k!.
- 2The five terms for k = 0 to 4 are 0.0821, 0.2052, 0.2565, 0.2138 and 0.1336, which sum to 0.8912.
- 1So P(X at least 5) = 1 minus 0.8912 = 0.1088, about an 11% chance in any one minute.
Key terms
- Sample space
- Every outcome an experiment can produce, collected into one set written S.
- Conditional probability
- The probability of A once B is known to have occurred, P(A and B) divided by P(B).
- Random variable
- A function that assigns a real number to each outcome in the sample space.
- Cumulative distribution function
- The function F(x) = P(X at most x), defined for every real x.
- Poisson distribution
- A count model with one parameter, the mean, for independent events at a constant rate.
- Monte Carlo simulation
- Repeated random draws from a model, used to approximate probabilities and distributions.
- Marginal distribution
- The distribution of one component of a random vector, found by summing or integrating out the others.
- Correlation coefficient
- Covariance divided by the product of standard deviations; it measures linear association between minus 1 and 1.
- Error propagation
- Approximating the variance of a function of uncertain inputs using its first derivatives and the input covariances.
- Random sample
- Independent random variables that all share one distribution, written X1 to Xn.
- Maximum likelihood estimator
- The parameter value that maximises the likelihood of the observed data.
- Likelihood ratio test
- A test that rejects the null when the maximised likelihood under it is small relative to the unrestricted maximum.
- Coverage probability
- The probability, over repeated samples, that an interval estimator contains the true parameter.
PHAS0083 FAQ
How hard is this statistics module for a physics graduate?
The difficulty is conceptual more than computational. The module moves from probability to frequentist inference in six teaching weeks, and the later chapters ask you to reason about repeated sampling, likelihoods and coverage rather than to recall formulas. The opening lecture says long derivations are not required, so the challenge is setting up the right model and statistic quickly.
How is the module assessed?
Weekly coursework carries 50% of the final grade and a final exam carries the other 50%. The coursework is five online quizzes weighted 5, 5, 10, 15 and 15 percent, due on Mondays from 19 October to 14 December. The final exam is in person and written. The module information does not give the exam date, which comes from the UCL timetable.
What will the final exam look like?
The opening lecture describes it as an in-person written exam worth 50%, says exam questions will closely resemble the coursework, and states that anything covered in the notes or lectures can be examined. Expect definition questions, short calculations of probabilities and expectations, likelihood set-ups, tests and intervals, rather than long proofs.
How do I enter answers in the coursework quizzes?
The quizzes use STACK, which checks algebraic answers. Keep the variable names and notation used in the question, type Greek letters by name without a backslash, and check the preview before submitting. Each coursework allows one attempt, so test unusual syntax on the dummy assessment first.
How much mathematics do I need?
You need confident calculus: sums and integrals for expectations, a change of variables with a Jacobian, partial derivatives for error propagation, and differentiating a log-likelihood. Matrix notation appears for covariance. The notes build every result from definitions, so revise by rederiving short results rather than memorising them.
Do I need to write code for this module?
Yes, for the coursework. The coursework papers include multistep questions with Python code parts that are checked automatically, alongside multiple-choice and numerical parts. The notes supply a Python script for every plot, and several chapters calibrate tests by simulation, so practise writing short simulations with a fixed seed and describing them precisely in words.
What should I revise first?
Start with the likelihood: writing it for a sample, maximising it and using a likelihood ratio. It links Chapters 7, 8 and 9 and the later weekly topics. Then secure distribution families and conditional probability, because every likelihood is built from them. Leave error propagation and density estimation for focused practice sessions.
Are past exam papers available here?
The available module materials contain the coursework papers and a dummy assessment, but no past final examination paper. The practice in this guide is original and follows the coursework question types, which the opening lecture says resemble the exam, without reproducing the university's questions.
Questions students ask in each PHAS0083 chapter
- Why do the notes say induction can never prove a model?
- Inductive reasoning generalises from the observations made so far, and a model that fits every past observation can still be contradicted by the next one. Data can raise confidence in one model and lower it in another, which is exactly what the likelihood later quantifies, but no finite set of agreeing results establishes the model as true. Chapter 01 · Statistical Inference and Probability Theory
- How do I pick the right family for a word problem?
- Look at what the experimental design fixes. A fixed number of independent trials points to the binomial, independent events in a fixed time or space window to the Poisson, the time until the next such event to the exponential, and a quantity built from many small independent effects to the normal. Equal likelihood on a range points to the uniform. Chapter 02 · Common Families of Distributions
- What makes a model hierarchical?
- A model qualifies as hierarchical when it is built from conditional steps and at least one variable is latent, meaning it is part of the model but never recorded. A detector that records only smeared energies of photons with randomly shifted true energies is an example. If every variable in the chain is recorded, the chain alone does not make it hierarchical. Chapter 03 · Monte Carlo Simulation and Density Estimates
- What is the quickest way to find the distribution of a sum?
- Introduce an auxiliary variable so the map is one-to-one, for instance U equal to X plus Y and V equal to Y. Transform the joint density with the Jacobian of the inverse map, then integrate out V. For discrete variables the same idea becomes a sum over all pairs that give each value of the total. Chapter 04 · Joint, Marginal and Conditional Distributions
- Should I propagate errors analytically or by simulation?
- Use the delta method when the function is close to linear over the spread of the inputs, because it is quick and needs only means and covariances. Use an exact change of variables or a Monte Carlo simulation when the function curves strongly, when a denominator can approach zero, or whenever the result will be quoted with high precision. Chapter 05 · Covariance, Correlation and Error Propagation
- What does convergence in probability actually guarantee?
- It guarantees that, for any tolerance you choose, the probability that the statistic lies within that tolerance of its limit tends to one as the sample grows. It does not say each additional observation moves the statistic closer, and a running mean can wander away for a while before settling. Chapter 06 · Random Samples and Convergence
- What should I check after setting the derivative of the log-likelihood to zero?
- Check that the stationary point is a maximum, using the second derivative or the shape of the curve, and that it lies inside the allowed parameter range. If the likelihood keeps increasing towards a boundary, as for a uniform upper limit or a non-negative signal rate, the maximum sits on that boundary instead. Chapter 07 · Point Estimation and Maximum Likelihood
- Where should I simulate when the null hypothesis is composite?
- The threshold must keep the level for every parameter value in the null. When the power function increases monotonically towards the alternative, the largest rejection probability inside the null occurs at its edge, so simulating at that boundary value is enough. Without monotonicity the whole null region would have to be scanned. Chapter 08 · Hypothesis Testing and Likelihood Ratio Tests
- Why can a fixed-length interval have zero confidence coefficient?
- Its coverage can depend on the parameter. For a uniform sample on zero to theta, an interval of the form maximum plus a fixed shift covers less and less often as theta grows, so the smallest coverage over all theta is zero. Scaling the interval with the maximum instead keeps the coverage constant. Chapter 09 · Interval Estimation and Coverage
How to study for the exam
Study PHAS0083 in the order the notes are written, because each backward-flow chapter reuses a forward-flow tool. In the first pass, make one page per distribution family with its support, parameter, mean, variance and the physical situation it describes, and derive each mean once from the definition.
In the second pass, work joint distributions by drawing the support region before integrating, because the limits of integration are where most marks are lost. For error propagation, write the gradient vector and the covariance matrix explicitly, then check the result against a quick simulation.
In the inference chapters, always write the likelihood for the whole sample, take the logarithm, and keep the parameter as a symbol until the end. For a test, state the null, the statistic, the rejection region and its size before any data are used. For an interval, say what is random and what is fixed, and phrase coverage as a property of the procedure.
Use the coursework papers as the template for timed practice, since the opening lecture says exam questions will be very similar, and enter algebra exactly as STACK expects. In the final fortnight, alternate one set of short definition questions with one set of likelihood-based problems each day.
Your AI Statistics tutor for PHAS0083
Stuck on a hard PHAS0083 question? Sia is AskSia’s AI Statistics tutor — ask any PHAS0083 Statistical Analysis of Data 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.