Sampling Designs
Every design here chooses at random in some way — except the last one. The question is
not whether randomness is used, but what it is applied to .
Simple random — everything equally likely. Fair, and scattered across
the whole site, so you cover a lot of ground.Stratified — split the site into its natural bands first, then sample
each at random in proportion to its size. Guarantees every band is represented.Cluster — pick a few whole groups at random and take
everything in them. Far less travelling.Multistage — pick a few groups, then a random sample
within each. Less travelling and less measuring.Convenience — take whatever is easy to reach. No randomness at all.
Strata and clusters are opposites
A stratum is a group you know about in advance and want to control: good
ones are uniform inside and different from each other. A cluster is a group
that is cheap to collect whole — and you do not choose what is inside it, you
get whatever the site's convenient units happen to contain.
That is why the setting matters. Each one offers different convenient units:
Cobble beach — a transect walked from the water up
(crossing every zone), or a quadrat : a one-metre square frame dropped
on the stones, so everything in it sits at one distance from the water. Field
ecologists use both, and you can switch between them here.Apple orchard — a tree . One ladder, one tree, all its
fruit. But a tree is one variety and its apples are all alike.Buried rocks — a hole , and only a hole. A patch at
depth would mean digging out everything above it and throwing the spoil away, which is
the opposite of convenient.
Watch how much you had to collect next to how close you got .
That trade — precision against effort — is the whole reason these designs exist.
Choose Show: Detailed for Run 500 of each design , which
compares the spread of the estimates. That is a standard-error question and belongs to a
later course; a first look at study design does not need it.
Close