Is cluster sampling is a type of stratified random sampling?

Answer 1

No. Cluster sampling and stratifed random sampling are different, though often confused. (They may, however, be used in conjunction in some sampling designs.) Both are types of random sampling.

STRATIFIED sampling involves identifying a variable that will break up your population into separate homogeneous groups (homogeneous in terms of the variable you are interested in). For example, suppose you want to know about the attitudes of kids about their future. Perhaps you have reason to believe this will change with time. If you collected a sample from high schools, you could stratify by grade, giving you 4 relatively homogeneous groups: freshmen, sophomores, juniors, seniors. Then, a common approach is to sample a similar number from each group.

Sometimes, though, separating the groups isn't so clear cut. Perhaps you want to stratify based on religion. You can't tell this from looking at a person. So perhaps you collect sample data and apply strata after the fact! This can be useful, but there are some statistical techniques that require equal (or nearly) sample sizes for the strata.

CLUSTER sampling involves breaking your population into fairly similarly sized groups called clusters (try googling MSE for an example). But now you want each cluster to contain a heterogeneous mix of individuals. Then, you take a random selection of these clusters and completely enumerate inside of those selected clusters. The problem with cluster sampling is that the cluster has now become your sample unit, instead of individuals which is what you probably hoped. This can be used for counting species, or just for contacting certain populations like apartment dwellers, nursing home residents, etc. The clusters could be apartment buildings in a city. So instead of taking a random sample of apartment dwellers, you would actually randomly select a few of the buildings and talk to everyone inside! Often, this is much more cost efficient. :)