Using summary statistics to estimate population characteristics; estimating the population mean from a sample; predicting population proportions; the effect of sample size on reliability and replication.

What this dot point is asking

Edexcel codes 2h.01 and 2h.03 require you to use summary statistics from a sample to estimate population characteristics, in particular to use a sample mean to estimate the population mean and to predict population proportions, and to know that sample size affects the reliability and replication of results. This is the heart of statistical inference: drawing conclusions about a whole population from a sample.

Using a sample to estimate the population

A well-chosen sample stands in for the population, so its summary statistics become estimates of the population's. Edexcel expects you to recognise, for example, that approximately half the population is expected to lie above the sample median, because the median splits the data in two. The reliability of any such estimate depends entirely on the sample being representative.

Estimating the population mean

The sample mean is the natural estimate of the population mean. If a sample of $50$ components has a mean mass of $12$ g, you estimate the mean mass of all components as $12$ g, and you can scale up to a total (for $20000$ components, an estimated total of $12 \times 20000 = 240{,}000$ g). This works because, for a random sample, the sample mean is centred on the population mean; it will not be exactly right, but it is the best single estimate.

Predicting population proportions

A sample proportion estimates the population proportion. If $130$ out of $200$ sampled voters say Yes, the estimated proportion is $\frac{130}{200} = 0.65$, and you predict that about $65\%$ of all voters would say Yes. To estimate a count in the population, multiply the proportion by the population size: $0.65 \times 50000 = 32{,}500$. This scaling-up is one of the most common inference tasks in the exam.

Sample size, reliability and replication

Code 2h.03 stresses that conclusions based on larger samples are generally more reliable, because random sampling variation has less effect on a big sample, so a repeat study is more likely to give a similar result. However, sample size does not cure bias: a large but unrepresentative sample (from a poor frame or a non-random method) still gives a biased estimate. So both a good method and an adequate size are needed.

This links directly to the quality-assurance idea that sample means vary less than individual values: averaging over more data smooths out the extremes, so a mean based on a larger sample is a tighter, more trustworthy estimate of the population mean. It is also why a single small sample should be treated with caution, and why repeating a study (replication) and getting a consistent answer strengthens confidence in the conclusion.

Estimating other characteristics

The same logic extends beyond the mean and a single proportion. A sample can be used to estimate the population median (about half the population lies above the sample median), the population range or spread, and the frequency of any category. In each case you treat the sample statistic as the best estimate of the matching population value, and where appropriate scale it up using the population size. Edexcel may give you a sample summary (a mean, a median, a set of class frequencies) and ask you to make a statement about the whole population, so practise turning a sample figure into a population estimate and stating clearly that it is an estimate, not an exact value.