There are explicit and implicit ways to stratify a sampling frame (divide the sampling units into smaller frames, essentially dividing the population into many strata). The purpose of stratification is to make sample estimates more precise and to allow oversampling if certain subgroups are of interest.
Explicit stratification separates the population into strata and then selects an independent sample from each stratum. It also weights the sample size in order to make the sample proportionate (e.g. allocate more villages to larger districts). This approach provides a better control of a subgroup’s sample size then the proportionate probability sampling approach because the allocated sample size (n) in each region are weighted by its own population distribution and therefore, the variability (variance) of the sample sizes in the subgroups will be reduced to zero. Using this approach, the sample size in each region exactly matches the % in the subgroup in the sample exactly matches that of the population because it is proportionate to the actual population.
* a sample size (n) of 10,000 was pre-determined
Before sampling is done, implicit stratified sampling procedure uses asorted list (e.g. villages within a district) and then takes a systematic sample from the sorted list using a fixed sampling interval and a random start. In other words, implicit stratification is a multistage geographic technique that combines elements of systematic sampling and stratified sampling. In order for each stratum to be representative, it is advised to choose a sampling interval that can cover the sorted sampling list.
* selected are highlighted in red
Sampling Interval = N / # To Be Selected = 4197/5 = 839.4
Random Start = 107.6
This approach may be quicker and easier to implement than explicit stratification for these reasons:
1) You only work with one list,
2) You only need to sort the file and apply the appropriate sampling intervals, and
3) It is not necessary to establish geographic strata and no sample size weighting is needed.
Implicit stratification can also circumvent the problem of small (or fractional, <1) sample sizes.
Under this approach, the % in the subgroup will closely resemble that of the population but may not be an exact match to the population's % because it unlike explicit stratification, it does not completely reduce the variability of the sample sizes in the subgroups through systematic sampling is reduced but not completely.
Both methods will yield the same results if sample size weighting is used in explicit stratification.
Depending on the research questions, both methods can be used in conjunction and implicit stratification can be nested within explicit stratification.