Multi-level Stratified Sampling
Also known as: hierarchical stratified sampling, nested stratified sampling, multilevel stratified design, stratified multilevel sampling
Multi-level stratified sampling applies stratification at two or more hierarchical levels of a nested population structure — for example, first stratifying geographic regions, then stratifying schools within each region, then stratifying classrooms within each school. This layered control over the composition of the sample at every level reduces variance and supports analysis at each level of the hierarchy, making it a powerful design for large-scale educational, epidemiological, and organizational surveys.
Read the full method
Sign in with a free account to read this section.
Method map
The neighbourhood of related methods — select a node to explore.
When to use it
Use multi-level stratified sampling when the target population has a well-defined hierarchical structure and you need precise estimates both overall and within subgroups at multiple levels — typical in national educational assessments (students nested in schools nested in districts), multi-site health surveys, and organizational research. It is especially valuable when resources allow sampling only a fraction of higher-level units and representation of important subgroups must be guaranteed at every tier. Do NOT use it when the population lacks a meaningful hierarchical structure, when frame information about strata at multiple levels is unavailable or unreliable, when the added administrative complexity is disproportionate to precision gains, or when you need a simple design for a small-scale exploratory study.
Strengths & limitations
- Controls representation at every level of the hierarchy, preventing accidental under-coverage of important subgroups.
- Reduces sampling variance compared with simple random sampling or single-level stratification, especially when strata are internally homogeneous.
- Supports valid multi-level analysis and subgroup estimation at each tier of the hierarchy.
- Allows different sampling fractions across strata to oversample rare or priority subgroups while maintaining design-based estimability.
- Transparent and reproducible — the stratification logic and inclusion probabilities can be fully documented and audited.
- Requires a complete and accurate sampling frame that maps units to strata at every level — often difficult to obtain and expensive to maintain.
- Design and implementation complexity increases substantially with each additional level of stratification.
- Differential sampling rates across strata necessitate weighting, which inflates variance if weight variation is large.
- Analysis must account for the complex survey design (clustering, stratification, unequal weights); naive standard errors from standard regression software are incorrect.
Frequently asked
How is multi-level stratified sampling different from multistage sampling?
Multistage sampling refers to the sequential selection of progressively smaller units (e.g., first select districts, then schools, then students). Stratification is an additional control applied at one or more stages to ensure proportional or targeted representation within defined subgroups. Multi-level stratified sampling combines both: it selects units in stages AND imposes stratification at multiple stages. Many large surveys use both features simultaneously.
Do I always need to stratify at every level?
No. You may stratify at only some levels where heterogeneity is greatest or where subgroup representation is most critical. Stratifying at the highest level (e.g., region) is almost always beneficial; stratifying at the lowest level (e.g., individual respondent) may be unnecessary if the final-stage sample within each cluster is already drawn randomly. Choose the levels where variance reduction or representation guarantees provide the most analytical value.
How do I handle the design weights in analysis?
Each sampled unit receives a design weight equal to the inverse of its overall probability of selection — computed as the product of selection probabilities across all levels. These weights must be applied in descriptive statistics and model-based analyses. Use software capable of complex-survey analysis (R's survey package, Stata's svy prefix, SAS proc surveymeans, SPSS Complex Samples) and specify the strata and cluster variables to obtain correct variance estimates.
What is the minimum strata size I should aim for?
A commonly cited rule of thumb is at least two sampled primary sampling units (PSUs) per stratum at the highest level to allow variance estimation. Strata with only one PSU make standard variance estimation impossible; such strata are often collapsed with adjacent strata before analysis. At lower levels, aim for strata with enough units that selection uncertainty is meaningful.
When should I prefer proportional versus optimal allocation?
Proportional allocation is simpler — sample proportional to stratum size — and is appropriate when within-stratum variances are similar. Optimal (Neyman) allocation samples more heavily from strata with greater variance and lower cost, minimizing overall sampling variance for a fixed budget. If you have reliable prior estimates of within-stratum variability and unit costs differ across strata, optimal allocation is more efficient; otherwise proportional allocation is robust and easier to justify.
Sources
- Cochran, W. G. (1977). Sampling Techniques (3rd ed.). John Wiley & Sons. ISBN: 978-0471162407
- Kish, L. (1965). Survey Sampling. John Wiley & Sons. ISBN: 978-0471489009
How to cite this page
ScholarGate. (2026, June 3). Multi-level Stratified Sampling. ScholarGate. https://scholargate.app/en/survey-methodology/multi-level-stratified-sampling
Which method?
Set this method beside its closest kin and read them side by side — the library lays the books on the table; the choice is yours.
- Cluster SamplingSurvey Methodology↔ compare
- Multi-level Cluster SamplingSurvey Methodology↔ compare
- Multistage SamplingSurvey Methodology↔ compare
- Proportional Stratified SamplingSurvey Methodology↔ compare
- Stratified SamplingSurvey Methodology↔ compare
- Systematic SamplingSurvey Methodology↔ compare