stratified sample

A stratified sample uses some characteristic such as age or gender to divide a population into sub-groups. This may be used to deliberately ensure that the groups are sampled in proportion to the prevalance in the population as a whole. Alternatively, some other criterion may be used to decide the subgroup samples; the statistics for each sub-group can then be weighted if the sub-groups have different proportions to the population proportion.

For example, suppose there are two subgroups with m₁ and m₂ individuals in the sample and n₁ and n₂ in the population as a whole. If we have measured means of μ₁ and μ₂ for the two subgroups our estimate of the overall population mean μ would be:
      ( μ₁×m₁/n₁  + μ₂×m₂/n₂ )   /   (n₁ + n₂)

It can be especially important to use stratified sampling when a subgroup has a very low prevalance, for example those with a rare illness, as simple random sample might either not have any members of the minority group at all, or have very small numbers hence making estimates about the subgroup unreliable.