Your starting point
Prerequisites: Use the ideas and vocabulary from lesson 16; revisit its worked examples if needed.
Definition: A population is the set of units a question concerns. A sample is the subset observed. Selection bias occurs when the way records enter the sample systematically distorts the intended inference. Random sampling and random assignment are different design choices.
Define who could be selected
A survey of people who attend an event cannot directly describe people who stayed away. The sampling frame is the practical list or process from which units can be selected; it may omit part of the intended population. Ask who is absent, who declines to respond and whether those differences relate to the outcome. A high response count does not answer these questions.
Bigger does not always repair bias
Collecting many more records from the same narrow source can estimate that source’s characteristics precisely while remaining unrepresentative of the population. Random variability and systematic selection problems are different. Increasing sample size often reduces random sampling variation under an appropriate design, but it does not automatically fix an excluded group or a misleading measurement.
Explain the design before the result
Document selection rules and relevant group composition. The academy simulator has a fully known synthetic population and deterministic selection methods, so it illustrates the consequences of a rule without claiming to estimate real-world uncertainty. An evenly spaced selection can work in the constructed ordering yet fail if real data have periodic patterns. Do not treat a convenient demonstration as a universal sampling recipe.
Two worked examples
A known population
Imagine 80 group-A records each valued 40 and 20 group-B records each valued 90. The population mean is (80×40 + 20×90) ÷ 100 = 50. Selecting only the first twenty A records yields a mean of 40. The arithmetic is correct for the sample, but the sample misses B entirely.
An optional poll
An online poll receives 1,000 responses from enthusiastic subscribers. A smaller carefully sampled survey might represent the target population better. The large poll’s size cannot establish that non-subscribers or nonrespondents share its views. State who responded rather than calling the result everybody’s opinion.
Where this becomes useful
A teacher evaluates a new study guide using feedback only from students who completed it. Learners who abandoned it may have different experiences, so the feedback can inform improvements without representing the whole class.
Make an attempt
Use Sampling Bias Simulator. Predict the result for first-record selection and evenly spaced selection at sizes 10, 20 and 50. Compare each sample with the known population mean and explain why the deterministic illustration is not a confidence interval.
Misconception check
Claim: A large sample must be representative.
Inspect the correction
Size and selection design address different problems. Inspect the frame, nonresponse and population coverage.
Independent application
Write a sampling plan for a fictional workshop survey. Define the population, frame, selection method and possible nonresponse. Describe one improvement to coverage and one limitation that would remain even with many responses.
Completion records practice and self-review. It is not certification of statistical competence.
