PHAS0083 Chap.1 Statistical Inference and Probability Theory
Statistical Inference and Probability Theory
Statistical Inference and Probability Theory sets up the language for the rest of the module.
It starts from why physical outcomes look random, either because initial conditions and environments are never known exactly or because quantum measurement is random in principle, and from the difference between deductive reasoning, which runs from a model to the data it predicts, and inductive reasoning, which runs from observed data back to models.
Four tasks organise inference: point estimation, interval estimation, hypothesis testing and goodness of fit, often chained on one data set. The probability half of the chapter defines a sample space, events and a probability function satisfying three axioms, then conditioning, Bayes' rule and independence.
Counting rules handle equally likely outcomes, with the choice between ordered and unordered selection and with or without replacement deciding the formula. Random variables reduce a large sample space to the quantity of interest, and their pmf, pdf and cdf carry the probabilities. A statistical model is then a family written with its parameter after a bar, the object every later chapter either simulates from or infers.
What this chapter covers
- 01
Two sources of randomness
- 02
Deductive and inductive reasoning
- 03
The four tasks of statistical inference
- 04
Sample spaces, events and the axioms
- 05
Conditional probability and Bayes' rule
- 06
Independence and counting rules
- 07
Random variables with pmf, pdf and cdf
Worked example · free
Find the purity of a muon tag with Bayes' rule
- 1Write the target as P(muon given tag) and list the inputs: prior 0.30, P(tag given muon) 0.98 and P(tag given pion) 0.03. This allocation is practice weighting, not a UCL marking scheme.
- 1Total probability of a tag: 0.98 times 0.30 plus 0.03 times 0.70, which is 0.294 plus 0.021, or 0.315.
- 1Bayes' rule: 0.294 divided by 0.315 gives 0.933.
- 1Interpret: about 93% of tagged particles are muons, so roughly one tag in fifteen is a pion.
Key terms
- Inductive reasoning
- Reasoning from observed data to the models that could have produced them, giving support but never proof.
- Event
- Any subset of the sample space, including the whole space and the empty set.
- Statistical independence
- The property P(A and B) = P(A)P(B), so knowing one event leaves the other's probability unchanged.
- Probability density function
- A function whose integral over an interval gives the probability of that interval for a continuous variable.
Statistical Inference and Probability Theory FAQ
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.
What is the difference between the forward and backward flow?
The forward flow starts from a model and predicts the ensemble of data sets it can produce, which is deduction and simulation. The backward flow starts from the observed data and asks which models or parameter values make those data likely, which is induction and the subject of inference.
Why can a probability density be larger than 1?
A density is a probability per unit of the variable, not a probability. A narrow normal curve with standard deviation 0.25 peaks near 1.6, yet the area under it over any interval stays between 0 and 1, and only those areas are probabilities.
Are disjoint events independent?
No. If two events with positive probability cannot happen together, learning that one occurred sets the probability of the other to zero, which is the opposite of independence. Independence requires the joint probability to equal the product of the separate probabilities.
Exam move
Start by writing out, in words, the forward and backward flow for one experiment you know, such as counting photons in a detector, and name which of the four inference tasks each question about it would be. Then drill the probability rules on small finite examples where you can list the sample space completely: two dice, a short sequence of coin tosses or a few detector channels.
For each problem write the event in set notation before computing, decide whether the complement is easier, and say whether outcomes are ordered and whether draws are replaced. Practise Bayes' rule by always computing the total probability of the conditioning event first.
For random variables, sketch the pmf or pdf and the cdf for each example, check that the cdf rises from 0 to 1, and remind yourself that a density value is not a probability. Finish by writing three statistical models in the course notation, with the parameter after a bar, because every later chapter starts from one.
Working through Statistical Inference and Probability Theory in PHAS0083? Sia is AskSia’s AI Statistics tutor — ask any PHAS0083 Statistical Inference and Probability Theory question and get a clear, step-by-step explanation grounded in how PHAS0083 is taught and assessed. Read this chapter free, then take your hardest questions to Sia.