Glossary of Key Terms
| Term | Definition |
| Stratified Sampling | A probability technique that divides the population into distinct, homogeneous subgroups (strata) and randomly samples within each. |
| Strata | The mutually exclusive subgroups into which the population is divided, based on a shared characteristic. |
| Stratification Variable | The characteristic (e.g., gender, year level, region) used to divide the population into strata. |
| Proportionate Stratified Sampling | A stratified approach where each stratum’s sample size matches its proportion in the population. |
| Disproportionate Stratified Sampling | A stratified approach where strata are sampled at different rates, often to ensure adequate representation of small subgroups. |
| Probability Sampling | Sampling in which every unit has a known, non-zero chance of selection. |
| Sampling Bias | Systematic error from unequal or non-random selection; reduced within strata under this method. |
| Internal Validity | The degree to which a study’s design supports confident conclusions about relationships within the sample. |
| External Validity | The degree to which findings generalize to the broader population. |
| Sample Size | The total number of units selected, distributed across strata. |
| Study Power | The probability of detecting a true effect, often improved within subgroups through stratification. |
| Effect Size | A standardized measure of the magnitude of a relationship or difference, which may vary by stratum. |
| Study Design | The overall plan guiding how a study is structured and conducted. |
| Research Question | The central question a study seeks to answer. |
| Research Objectives | Specific, measurable goals that operationalize the research question. |
Key Takeaways
- Stratified sampling divides a population into homogeneous strata before applying random sampling within each, improving precision compared to simple random sampling at the same sample size.
- It is especially valuable in quantitative research when subgroup comparisons are part of the research question or research objectives.
- Done well, it strengthens both internal and external validity by ensuring key subgroups are adequately represented.
- Sample size and study power calculations must account for each stratum individually, not just the overall sample.
What Is Stratified Sampling?
Definition
Stratified sampling is a probability technique in which the population is first divided into mutually exclusive, internally homogeneous subgroups called strata (e.g., by gender, age group, or year of study), and a random sample is then drawn from each stratum, either proportionately or disproportionately.
Where It Fits in Study Design
Stratified sampling is common in quantitative study designs, particularly surveys and comparative studies, where the research question involves comparing defined subgroups. It can also support the quantitative phase of mixed-methods designs.
Purpose: When and Why to Use It
- Used when the research question requires reliable comparisons across known subgroups.
- Appropriate when certain subgroups are small and might be underrepresented by simple random sampling.
- Improves precision and statistical efficiency compared to simple random sampling at an equivalent sample size.
- Useful when prior information about relevant population characteristics (the stratification variable) is available.
Fit with Quantitative, Qualitative, and Mixed-Methods Research
| Approach | Typical Role of Stratified Sampling | Example |
| Qualitative | Occasionally used to ensure diverse case representation across known subgroups before purposive selection | Selecting interview participants from each academic department before purposive criteria are applied |
| Quantitative | Common method for ensuring adequate, precise representation of subgroups in survey or experimental research | Surveying proportionate numbers of students from each year level |
| Mixed Methods | Supports a representative quantitative phase, later complemented by qualitative follow-up within strata | Surveying stratified groups by department, then interviewing representatives from each |
How It Works
Step-by-Step Process
- Define the population and the research question, including any subgroup comparisons of interest.
- Identify a relevant stratification variable (e.g., gender, region, year level).
- Divide the population into mutually exclusive, exhaustive strata based on that variable.
- Decide between proportionate or disproportionate allocation across strata.
- Randomly sample within each stratum using simple random sampling or systematic sampling.
- Combine stratum samples into the final overall sample, applying weights if disproportionate allocation was used.
Types and Variations
| Type | Description |
| Proportionate Stratified Sampling | Sample size within each stratum mirrors that stratum’s share of the total population. |
| Disproportionate Stratified Sampling | Sample sizes differ across strata, often oversampling smaller groups to allow meaningful subgroup analysis. |
| Post-Stratification | Stratification applied after data collection by re-weighting an already-collected sample to match known population proportions. |
Strengths and Limitations
Strengths
- Increases precision and can improve study power for subgroup comparisons compared to simple random sampling.
- Ensures adequate representation of important, sometimes small, subgroups.
- Reduces sampling error when strata are genuinely homogeneous internally.
- Supports both overall population estimates and detailed subgroup analysis within a single study design.
Limitations
- Requires accurate prior information to define meaningful strata.
- More complex to design, execute, and analyze than simple random sampling.
- Choosing an irrelevant or poorly defined stratification variable reduces or eliminates the precision benefits.
- Disproportionate allocation requires careful weighting during analysis to avoid biased overall estimates.
Effect on Internal and External Validity
| Validity Type | Typical Impact |
| Internal Validity | Often strengthened, since ensuring subgroup representation supports more credible within-sample comparisons and conclusions. |
| External Validity | Typically strong, often stronger than simple random sampling, because stratification ensures the sample mirrors known population subgroup proportions. |
Sample Size, Effect Size, and Study Power
Sample size planning for stratified sampling must consider both the overall sample size and the sample size needed within each stratum to achieve adequate study power.
- Each stratum may require its own minimum sample size to detect a meaningful effect size for subgroup comparisons.
- Smaller strata often need disproportionate oversampling to reach adequate study power for that subgroup.
- Overall study power can be higher than with simple random sampling of the same total size, since stratification reduces between-group variability.
- Statistical weighting is needed during analysis whenever disproportionate allocation is used, to avoid distorting overall population estimates.
Guidance by Academic Level
For Undergraduate Students
- Choose a stratification variable that is clearly relevant to your research question, such as class year or major.
- Proportionate stratified sampling is usually simpler to explain and execute for a class project than disproportionate allocation.
- Clearly show in your methods section how you divided the population into strata and how many participants came from each.
- Ask your instructor whether subgroup comparisons are expected, since this affects whether disproportionate sampling is worth the added complexity.
For Graduate Students
- Justify your choice of stratification variable using theory or prior research relevant to your research question.
- Report stratum-specific sample sizes and any disproportionate allocation decisions, along with the weighting approach used in analysis.
- Conduct a power analysis for each key subgroup comparison, not just for the overall sample.
- Discuss how stratified sampling strengthens your study’s internal and external validity relative to simpler alternatives.
Implementation Checklist
- Define the research question and any subgroup comparisons of interest.
- Select a meaningful, well-justified stratification variable.
- Divide the population into mutually exclusive strata.
- Decide on proportionate or disproportionate allocation.
- Randomly sample within each stratum and apply weights if needed.
- Report stratification approach and sample sizes per stratum transparently.
Common Mistakes to Avoid
- Choosing a stratification variable unrelated to the research question.
- Failing to apply statistical weights after disproportionate allocation.
- Overlapping or incomplete strata that violate the requirement of mutually exclusive groups.
- Treating stratified sampling as equivalent to cluster sampling, when the two methods serve different purposes.
Frequently Asked Questions
How is stratified sampling different from cluster sampling?
Stratified sampling divides the population into homogeneous subgroups and samples within every stratum, increasing precision. Cluster sampling divides the population into heterogeneous groups (clusters) and randomly selects entire clusters, primarily to reduce cost and logistical burden, often at some cost to precision.
When should I use disproportionate rather than proportionate stratified sampling?
Use disproportionate allocation when a subgroup is small relative to the total population but still requires a large enough sample size to achieve adequate study power for meaningful subgroup analysis.
Do I need to weight my data after stratified sampling?
Weighting is required after disproportionate stratified sampling to produce unbiased overall population estimates. Proportionate stratified sampling generally does not require weighting, since stratum sample sizes already mirror population proportions.
Can stratified sampling improve my study’s effect size estimates?
Stratified sampling does not change the true effect size in the population, but by reducing variability within strata, it can produce more precise and reliable estimates of effect size and improve overall study power.
This article was originally published on May 28, 2024, and updated on June 16, 2026.
