{
  "id": "stratified-sampling",
  "title": "Stratified Sampling",
  "lang": "en",
  "source_lang": "en",
  "summary": "Stratified sampling divides a population into nonoverlapping groups and samples within each group to improve precision, ensure subgroup coverage, or meet operational requirements.",
  "category": "mathematics",
  "subcategory": "probability-statistics",
  "type": "concept",
  "aliases": [
    "Stratified random sampling"
  ],
  "generated_at": "2026-11-04",
  "model": "pro/gpt-6.1-sol",
  "url": "https://aiwiki.page/en/stratified-sampling/",
  "translations": {
    "zh-Hans": {
      "url": "https://aiwiki.page/zh/stratified-sampling/",
      "json": "https://aiwiki.page/zh/stratified-sampling.json"
    }
  },
  "sections": [
    {
      "heading": "",
      "text": "Stratified sampling is a method of random sampling in which a population is divided into mutually exclusive, collectively exhaustive groups called strata, and a sample is selected separately within each stratum. In statistics, it is used to improve the precision of population estimates, obtain adequate information about particular subgroups, or organize data collection. Population estimates combine the stratum results using weights that reflect the sampling design; an unweighted analysis is generally inappropriate when different strata have different sampling fractions."
    },
    {
      "heading": "Design and rationale",
      "text": "A stratification is a partition of the population: every population unit belongs to exactly one stratum. Strata may be defined by geographic region, business size, age group, land-cover type, or combinations of characteristics known before sampling.\n\nA basic stratified design involves four steps:\n\n- Define the target population and a sampling frame that identifies its units.\n- Assign each unit to a stratum.\n- Allocate a sample size to each stratum.\n- Select units independently across strata using a specified probability-sampling method.\n\nThe most common elementary design uses simple random sampling without replacement within every stratum. More complex designs may use unequal selection probabilities or multiple sampling stages within strata.\n\nThe main precision advantage arises when units within each stratum are relatively similar for the quantity being measured, while stratum means differ. Separate sampling fixes the number of observations drawn from each group, reducing variation caused by an accidentally unbalanced sample composition.\n\nStratification can also serve goals other than improving an overall estimate. A small population subgroup may need a large sampling fraction to support a useful subgroup estimate. Administrative regions may require separate results, or different strata may require different collection methods."
    },
    {
      "heading": "Estimation and sampling weights",
      "text": "Suppose a finite population contains \\(N\\) units divided into \\(H\\) strata. Let\n\n- \\(N_h\\) be the population size of stratum \\(h\\);\n- \\(n_h\\) be its sample size;\n- \\(W_h=N_h/N\\) be its population share;\n- \\(f_h=n_h/N_h\\) be its sampling fraction;\n- \\(\\bar y_h\\) be its sample mean.\n\nThe total sample size is \\(n=\\sum_{h=1}^{H}n_h\\). Under simple random sampling within strata, an estimator of the population mean is\n\n\\[\\hat{\\bar Y}_{\\mathrm{st}} =\\sum_{h=1}^{H}W_h\\bar y_h.\\]\n\nAn estimator of the population total is\n\n\\[\\hat T_{\\mathrm{st}} =\\sum_{h=1}^{H}N_h\\bar y_h =N\\hat{\\bar Y}_{\\mathrm{st}}.\\]\n\nThese estimators are unbiased under the stated sampling design: their expected values over repeated sampling equal the corresponding population quantities.\n\nWithin stratum \\(h\\), each unit has inclusion probability\n\n\\[\\pi_{hi}=\\frac{n_h}{N_h},\\]\n\nand therefore the base sampling weight is\n\n\\[d_{hi}=\\frac{1}{\\pi_{hi}}=\\frac{N_h}{n_h}.\\]\n\nEach sampled unit represents \\(N_h/n_h\\) population units. Because these weights sum to \\(N\\), the weighted sample mean equals \\(\\hat{\\bar Y}_{\\mathrm{st}}\\).\n\nThe unweighted sample mean instead weights each stratum by its sample share, \\(n_h/n\\). It agrees with the stratified estimator when those shares equal the population shares, as in exact proportional allocation. Otherwise, it generally estimates a differently weighted quantity.\n\nFor a population proportion, the same formulas apply by coding each observation with an indicator function equal to 1 when the characteristic is present and 0 otherwise."
    },
    {
      "heading": "Variance and uncertainty",
      "text": "Let the finite-population variance within stratum \\(h\\) be\n\n\\[S_h^2 =\\frac{1}{N_h-1} \\sum_{i=1}^{N_h}(Y_{hi}-\\bar Y_h)^2,\\]\n\nwhere \\(\\bar Y_h\\) is the stratum population mean.\n\nIf samples are selected independently across strata, the variance of the stratified mean estimator is\n\n\\[\\operatorname{Var}(\\hat{\\bar Y}_{\\mathrm{st}}) = \\sum_{h=1}^{H} W_h^2(1-f_h)\\frac{S_h^2}{n_h}.\\]\n\nThe factor \\(1-f_h\\) is the finite-population correction. It accounts for sampling without replacement and becomes important when a substantial proportion of a stratum is observed. A completely enumerated stratum, with \\(n_h=N_h\\), contributes no sampling variance.\n\nAn unbiased variance estimator replaces \\(S_h^2\\) with the sample variance \\(s_h^2\\):\n\n\\[\\widehat{\\operatorname{Var}}(\\hat{\\bar Y}_{\\mathrm{st}}) = \\sum_{h=1}^{H} W_h^2(1-f_h)\\frac{s_h^2}{n_h}.\\]\n\nIts square root is the estimated standard error. Under suitable large-sample conditions, an approximate confidence interval is\n\n\\[\\hat{\\bar Y}_{\\mathrm{st}} \\pm z_{1-\\alpha/2} \\sqrt{\\widehat{\\operatorname{Var}}(\\hat{\\bar Y}_{\\mathrm{st}})},\\]\n\nwhere \\(z_{1-\\alpha/2}\\) is a standard normal-distribution quantile.\n\nSmall samples, highly skewed outcomes, and complex within-stratum designs may require different approximations. A noncensus stratum containing only one sampled unit does not provide a direct within-stratum sample variance estimate, so additional assumptions or specialized variance procedures are needed."
    },
    {
      "heading": "Allocation of the sample",
      "text": "Allocation determines how the total sample is distributed across strata. Different allocations serve different objectives.\n\nProportional allocation\n\nProportional allocation sets\n\n\\[n_h=n\\frac{N_h}{N}=nW_h.\\]\n\nIgnoring integer rounding, all strata have the same sampling fraction and all sampled units have the same base weight. The design is then self-weighting.\n\nThis allocation preserves population shares in the sample but may provide too few observations for small subgroups.\n\nEqual allocation\n\nEqual allocation assigns approximately \\(n/H\\) observations to every stratum. It can support comparisons among strata, but it oversamples small strata and undersamples large ones relative to their population shares.\n\nEqual allocation is not generally optimal for estimating the overall population mean. Its usefulness depends on the importance attached to subgroup estimates.\n\nNeyman allocation\n\nFor a fixed total sample size, equal per-unit costs, and simple random sampling within strata, the allocation minimizing the variance of the overall mean is\n\n\\[n_h = n\\frac{N_hS_h}{\\sum_{k=1}^{H}N_kS_k}.\\]\n\nThis is called Neyman allocation, after Jerzy Neyman, whose 1934 work helped establish the design-based theory of stratified sampling.\n\nThe rule assigns more observations to strata that are larger or have greater within-stratum standard deviation. Since the \\(S_h\\) values are usually unknown in advance, planning may use pilot studies, earlier surveys, or related auxiliary information.\n\nAllocation with unequal costs\n\nIf the cost per sampled unit in stratum \\(h\\) is \\(c_h\\), a variance-minimizing allocation under a fixed variable-cost budget has the form\n\n\\[n_h\\propto\\frac{N_hS_h}{\\sqrt{c_h}}.\\]\n\nOther things equal, expensive strata receive fewer observations than under Neyman allocation.\n\nThese formulas describe continuous optima. Actual allocations require integer sample sizes, cannot exceed stratum population sizes, and may impose minimum samples for subgroup reporting. With several important outcomes or precision constraints, allocation becomes a mathematical optimization problem rather than a single universal rule."
    },
    {
      "heading": "Numerical example",
      "text": "Consider a population of 10,000 businesses:\n\nStratum | Population size | Sample size | Sample mean revenue\nSmall businesses | 9,000 | 100 | 100,000 currency units\nLarge businesses | 1,000 | 100 | 1,000,000 currency units\n\nThe estimated population mean revenue is\n\n\\[\\hat{\\bar Y}_{\\mathrm{st}} = 0.9(100{,}000)+0.1(1{,}000{,}000) = 190{,}000.\\]\n\nThe unweighted sample mean would be 550,000 because the two strata contribute equal numbers of observations. That calculation gives large businesses half the weight, although they make up only one-tenth of the population.\n\nThe base weights are 90 for each sampled small business and 10 for each sampled large business. Estimated total revenue is\n\n\\[\\hat T_{\\mathrm{st}} = 10{,}000(190{,}000) = 1.9\\text{ billion currency units}.\\]\n\nThe sample means alone do not determine the precision of this estimate; within-stratum variances are also required."
    },
    {
      "heading": "Choosing strata and assessing efficiency",
      "text": "Effective stratification variables are available for the sampling frame before selection and are informative about important survey outcomes. A variable strongly associated with one outcome may be weakly associated with another, so a stratification suitable for estimating income may not be equally effective for estimating travel behavior.\n\nStrata need not have equal sizes or correspond to socially meaningful groups. Continuous variables, such as business employment or farm area, can be divided into intervals. Very large or influential units may form a certainty stratum, in which every unit is selected.\n\nPrecision gains depend on both the boundaries and the allocation. Under proportional allocation, stratification reduces the contribution of between-stratum differences to sampling variability; with negligible sampling fractions, this gives a clear variance advantage when stratum means differ. However, an inefficient disproportionate allocation can produce a larger variance than simple random sampling of the same total size.\n\nA common efficiency measure is the design effect:\n\n\\[\\operatorname{DEFF} = \\frac{\\operatorname{Var}(\\hat\\theta_{\\mathrm{design}})} {\\operatorname{Var}(\\hat\\theta_{\\mathrm{SRS}})},\\]\n\nwhere the denominator refers to an appropriate simple-random-sampling benchmark. A value below 1 indicates lower sampling variance for that estimator and comparison."
    },
    {
      "heading": "Related methods",
      "text": "Cluster sampling\n\nIn cluster sampling, groups of units are selected, and observations are collected from all or some units within the selected groups. In a conventional stratified design, by contrast, a sample is drawn from every stratum.\n\nStratification often seeks internally homogeneous groups to improve precision. Cluster sampling often uses naturally occurring groups, such as schools or neighborhoods, to reduce collection costs. The two methods can be combined: clusters may be selected separately within geographic strata.\n\nPost-stratification\n\nPost-stratification adjusts weights after sampling so that weighted counts agree with known population totals for specified groups. Unlike design stratification, it does not fix subgroup sample sizes before selection and cannot ensure that every subgroup appears in the sample.\n\nBoth methods use auxiliary population information, but their sampling properties and variance calculations differ.\n\nQuota sampling\n\nQuota sampling may also set target numbers for subgroups, but selection within groups need not be random. Meeting subgroup quotas alone does not establish known inclusion probabilities or justify the design-based formulas used for stratified random sampling.\n\nStratified randomization\n\nIn experimental design, treatment assignment may be randomized separately within strata to balance treatment groups on selected characteristics. This concerns treatment assignment rather than selection from a population. Stratified sampling and stratified randomization can occur in the same study but address different sources of variation."
    },
    {
      "heading": "Applications and limitations",
      "text": "Stratified sampling is widely used in household surveys, business statistics, agricultural surveys, environmental monitoring, and quality assessment. It supports both overall estimates and estimates for reporting domains. In Monte Carlo methods, an analogous technique partitions a simulation domain and samples within its parts to reduce numerical estimation variance.\n\nIts main practical limitation is the need for adequate frame information. Missing units, outdated classifications, or incorrect stratum assignments can undermine coverage and efficiency. Creating many strata can also increase administrative complexity and leave too few observations in individual strata for stable variance estimation.\n\nStratification does not by itself remove measurement error, nonresponse, or selection bias caused by departures from the intended design. Weight adjustments for nonresponse and calibration may be needed, and uncertainty calculations must reflect the actual selection and adjustment procedures. Treating a complex stratified sample as an ordinary independent, identically distributed sample can give incorrect standard errors even when the point estimate uses suitable weights."
    }
  ],
  "links": [
    "random-sampling",
    "statistics",
    "set-partition",
    "sample-size",
    "estimator",
    "estimator-bias",
    "expected-value",
    "sample-mean",
    "indicator-function",
    "variance",
    "standard-error",
    "confidence-interval",
    "normal-distribution",
    "standard-deviation",
    "mathematical-optimization",
    "experimental-design",
    "monte-carlo-method",
    "selection-bias"
  ],
  "sources": [],
  "license": "https://creativecommons.org/licenses/by/4.0/"
}