Math and Stats

Statistics

Sampling Methods on a Map

Can a sample give a precise estimate of the wrong neighborhood?

Draw a sample from the whole town. Switch to the riverfront convenience route, then compare 500 samples from each design.

Design a canopy survey

Each square on the map is an equal-area block in a fictional town. Its number gives the percentage covered by tree canopy. Estimate the town mean from a sample of blocks.

Town canopy map

Selected blocks have a double outline. Read each block as canopy cover in percent.

Sample addresses and readings

Teaching notes and sampling assumptions

Every sample uses distinct blocks. Simple random sampling gives each possible set of the chosen size equal probability. Stratified sampling draws equal numbers independently from three equally sized districts. In both cases the sample mean estimates the town mean without bias.

For the convenience design the survey team visits only the riverfront because it is nearby. We model variation among visits by drawing blocks at random within that restricted frame. Its expected mean is the riverfront mean, even though the target is the entire town. This is a model of undercoverage; other convenience samples may involve additional sources of bias.

Compare sample sizes 3 and 15. More observations reduce the spread in all three designs here. Which designs center on the town value? In this town, variation within districts is smaller than variation between districts. This is why stratifying helps.

Each estimate is the unweighted mean of selected block percentages. Equal block areas and proportional allocation justify those weights. The exact sampling SD includes the correction for drawing without replacement. Simulation SD uses the sample standard deviation of 500 estimates. Each draw or comparison uses a fresh seed. Select Fix seed to reproduce a sample or compare designs under the same settings.