Let’s assume that you want to make a survey of individuals in Regina who live in apartments. You could make a list of every individual who lives in all apartment blocks in the city. This would be your sample frame. This can be done, but it is time consuming and expensive.
Instead, you could make a sample of apartment buildings. For example, you could sample 20 apartment buildings in Regina. You could make a list of all adults in those buildings, and then randomly sample 80 people from this list.
Instead of making a list of everyone in the selected buildings, you might decide to sample 10% of the individuals in each building. This can lead to an under-representation of individuals living in small apartment buildings. A building that has 200 units would have 20 units sampled. What would you do with a building that only has 4 apartments? 10% of 40 is .4, which rounds down to zero. If there are a lot of small apartment buildings, there is a good chance you might eliminate a large proportion of renters from your sample. This is not a problem if the number of individuals in each cluster is large.
A poor method is to create your sample of apartment buildings, and then sample 4 people from each. You still will obtain a sample of 80 people, but you will over-represent those who live in small apartment buildings, and under-represent those who live in large apartment buildings.
For example, you may only sample 2 buildings. One building may have 200 units, the other may have 4 units. If you sample 4 people from each building, you will sample 4/200 or 2% of the units in the large building, and all of the units in the small building.
School children. You could first sample a cluster of schools, and from within these schools you could randomly sample children.
Hospitals. There are a number of hospitals in Saskatchewan. Some are large, and some are quite small.