PHAS0083 Chap.6 Random Samples and Convergence
Random Samples and Convergence
Random Samples and Convergence turns the module from the forward flow to the backward flow. Repeated measurements are modelled as a random sample, independent and identically distributed variables whose joint density is the product of identical marginals with a shared unknown parameter.
A statistic is a quantity computed from the observations alone that does not use the unknown parameter; it is a random variable whose distribution the population determines. The notes describe three layers, population, statistic and observed value, with simulation running down the layers and inference running up.
The sample mean is unbiased with variance sigma squared over n, and the sample variance divides by n minus 1 to remove the bias created by measuring deviations from the sample mean. Convergence is then made precise: sequences of statistics, convergence in probability, the weak law of large numbers and the central limit theorem, which makes the sample mean approximately normal whatever the population.
Finally Monte Carlo is revisited in two roles: one large set of draws to learn a population, and an ensemble of simulated data sets of the real size to learn the distribution of a statistic.
What this chapter covers
- 01
Random sample and population
- 02
Statistics and their distributions
- 03
Sample mean and its variance
- 04
Sample variance and the n minus 1 divisor
- 05
Sample correlation
- 06
Law of large numbers
- 07
Central limit theorem and simulated ensembles
Worked example · free
Use the central limit theorem for a mean of measurements
- 1The mean has standard deviation 3/sqrt(25) = 0.6. This allocation is practice weighting, not a UCL marking scheme.
- 1By the central limit theorem, the mean is approximately normal with mean 50 and standard deviation 0.6.
- 2Standardise: z = (51 - 50)/0.6 = 1.67, so the upper tail is 1 - Phi(1.67) = 0.048.
Key terms
- Population
- The distribution from which each observation in a random sample is drawn.
- Sample variance
- The sum of squared deviations from the sample mean divided by n minus 1.
- Convergence in probability
- The property that the probability of a statistic being within any fixed distance of its limit tends to one.
Random Samples and Convergence FAQ
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.
Why does the sample variance divide by n minus 1?
Deviations measured from the sample mean are on average smaller than deviations from the true mean, because the sample mean sits in the middle of the data. The expected sum of squares is n minus 1 times sigma squared, so dividing by n minus 1 makes the estimator unbiased.
What is the difference between the sample variance and the variance of the mean?
The sample variance estimates how much single measurements spread around their mean. The variance of the mean is that spread divided by n, and describes how much the average itself would vary between repeated experiments of the same size.
Does the central limit theorem need a normal population?
No. For independent and identically distributed observations with finite variance, the standardised sample mean tends to a standard normal whatever the population shape. How large n must be depends on the skewness of the population.
Exam move
Begin by writing the joint density of a random sample for three families and identifying the shared parameter in each. Then practise deciding whether a given quantity is a statistic, which fails as soon as it involves an unknown parameter. Derive the expectation of the sample variance once, line by line, so the n minus 1 divisor becomes a result rather than a rule.
For convergence, run a running mean on simulated data and watch it settle, then simulate many samples of a fixed size from a skewed population and compare the histogram of their means with the normal curve predicted by the central limit theorem. Practise sample-size questions in both directions, and always state which quantity grows in a simulation: the number of draws or the number of data sets of the real size.
Working through Random Samples and Convergence in PHAS0083? Sia is AskSia’s AI Statistics tutor — ask any PHAS0083 Random Samples and Convergence 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.