{
  "id": "repeatability",
  "title": "Repeatability",
  "lang": "en",
  "source_lang": "en",
  "summary": "Repeatability is measurement precision under specified, closely controlled conditions, usually involving the same procedure, operator, equipment, location, and a short time interval.",
  "category": "science",
  "subcategory": "physics",
  "type": "concept",
  "aliases": [
    "Measurement repeatability"
  ],
  "generated_at": "2026-11-05",
  "model": "pro/gpt-6.1-sol",
  "url": "https://aiwiki.page/en/repeatability/",
  "translations": {
    "zh-Hans": {
      "url": "https://aiwiki.page/zh/repeatability/",
      "json": "https://aiwiki.page/zh/repeatability.json"
    }
  },
  "sections": [
    {
      "heading": "",
      "text": "Repeatability is the closeness of agreement among repeated measurements made under a specified set of conditions that are kept as nearly unchanged as practicable. In metrology, it is a form of measurement precision, normally assessed using the same measurement procedure, operator, measuring system, operating conditions, and location over a short period. Repeatability describes the dispersion of results, not their closeness to a true or reference value: measurements can be highly repeatable while consistently wrong."
    },
    {
      "heading": "Definition and repeatability conditions",
      "text": "A repeatability assessment concerns repeated measurements of the same object, or of similar objects whose differences are negligible for the purpose of the assessment. The quantity intended to be measured is the measurand. Both the measurand and the measurement procedure must be sufficiently well defined for the observed variation to have a clear interpretation.\n\nTypical repeatability conditions include:\n\n- the same measurement procedure;\n- the same operator or operators;\n- the same measuring system;\n- the same location and operating conditions;\n- measurements repeated within a short period.\n\n“Short” is relative to the process being studied. It means a period over which relevant changes, such as instrument drift, reagent deterioration, or environmental variation, are sufficiently restricted. It does not imply a universal number of minutes or hours.\n\nRepeatability does not require identical results. Even under closely controlled conditions, readings can vary because of electronic noise, small differences in positioning, sample handling, environmental fluctuations, or the finite resolution of an instrument. Its assessment quantifies the resulting spread.\n\nThe scope of the repeated operation is crucial. Repeatedly reading one prepared solution assesses a narrower part of an analytical procedure than independently preparing and measuring several portions of a material. Both can provide useful repeatability estimates, but they describe different processes."
    },
    {
      "heading": "Distinction from related concepts",
      "text": "Precision, trueness, and accuracy\n\nPrecision concerns agreement among repeated results. Repeatability is precision under a particular, restrictive set of conditions.\n\nMeasurement trueness concerns agreement between the average of a very large number of replicate measurements and a reference value. A persistent displacement from that reference is associated with systematic measurement error and may be estimated as bias.\n\nMeasurement accuracy concerns closeness to a true quantity value and is related to both precision and trueness. It is not interchangeable with repeatability.\n\nFor example, a balance might repeatedly report approximately 100.20 g for a reference mass whose assigned value is 100.00 g. A very small spread indicates good repeatability, but the displacement indicates a problem with trueness. Calibration can establish the relationship between instrument indications and reference values; it does not, by itself, eliminate all sources of variation.\n\nIntermediate precision and reproducibility\n\nThe distinction between repeatability and broader precision conditions depends on which factors are allowed to change.\n\nConcept | Typical conditions\nRepeatability | Same procedure, operators, measuring system, location, and short interval\n[[intermediate-precision | Intermediate precision]]\n[[reproducibility | Reproducibility]]\n\nIntermediate precision often describes the performance of a procedure within a laboratory over time. Reproducibility commonly concerns comparisons between laboratories.\n\nThese terms are not used uniformly across all scientific disciplines. In computational research, for example, reproducibility often refers to obtaining consistent computational results using the original data and code, while replicability may refer to obtaining consistent findings with new data. Such conventions should not be assumed to match metrological terminology."
    },
    {
      "heading": "Quantitative assessment",
      "text": "Repeatability standard deviation\n\nFor \\(n\\) repeated measurements \\(x_1,\\ldots,x_n\\), the sample mean is\n\n\\[\\bar{x}=\\frac{1}{n}\\sum_{i=1}^{n}x_i.\\]\n\nA common estimate of repeatability is the sample standard deviation,\n\n\\[s_r=\\sqrt{\\frac{\\sum_{i=1}^{n}(x_i-\\bar{x})^2}{n-1}}.\\]\n\nThe subscript \\(r\\) identifies repeatability conditions. The quantity \\(s_r\\) has the same units as the measured quantity. Its square estimates the variance attributable to the repeated measurement process under those conditions.\n\nA smaller standard deviation indicates tighter clustering, but it is meaningful only alongside the measurand, measurement range, procedure, and study conditions. A value established at one concentration or operating point need not apply across an instrument’s full range.\n\nFor positive quantities measured on a ratio scale, relative dispersion is sometimes reported as a repeatability coefficient of variation:\n\n\\[\\mathrm{CV}_r=\\frac{s_r}{\\bar{x}}\\times100\\%.\\]\n\nThis percentage can be useful when variability scales with measurement magnitude. It becomes unstable near a mean of zero and is generally inappropriate for scales with an arbitrary zero, such as Celsius temperature.\n\nRepeatability limit\n\nA repeatability limit, conventionally denoted \\(r\\), describes a difference that the absolute difference between two single test results is expected not to exceed with a specified probability under repeatability conditions.\n\nFor independent, approximately normally distributed results with a common repeatability standard deviation \\(\\sigma_r\\), their difference has standard deviation \\(\\sqrt{2}\\sigma_r\\). A two-sided coverage probability of approximately 95% therefore gives\n\n\\[r\\approx1.96\\sqrt{2}\\,\\sigma_r\\approx2.8\\sigma_r.\\]\n\nIn practice, an estimated \\(s_r\\) is often substituted:\n\n\\[r\\approx2.8s_r.\\]\n\nThis conventional approximation assumes that the standard deviation is adequately estimated and that the distributional assumptions are suitable. An exact prediction procedure based on a small dataset must account for uncertainty in that estimate.\n\nThe repeatability limit is not a confidence interval for the mean, an accuracy specification, or a limit on the difference between a result and a reference value. The definition of a “test result” also matters: a result consisting of an average of several readings has different variability from one individual reading.\n\nPooling estimates\n\nWhen several groups of measurements share a common repeatability variance, a pooled estimate can be calculated from their within-group variances:\n\n\\[s_{r,\\mathrm{pooled}}^2= \\frac{\\sum_{j=1}^{k}(n_j-1)s_j^2} {\\sum_{j=1}^{k}(n_j-1)},\\]\n\nwhere \\(n_j\\) and \\(s_j\\) are the number of observations and standard deviation in group \\(j\\).\n\nPooling is not appropriate merely because measurements share units. Variability may change with concentration, material, or operating range. In those circumstances, separate estimates or a model of changing variance may be more informative."
    },
    {
      "heading": "Designing a repeatability study",
      "text": "A repeatability study identifies the level at which observations are genuinely repeated and the sources of variation included in the estimate.\n\nImportant design features include:\n\n- Definition of the result: a single reading, an average, or the output of an entire procedure.\n- Independent repetition: whether the instrument, specimen, or preparation is reset between measurements.\n- Material stability: whether repeated testing changes the object being measured.\n- Environmental control: the permitted ranges of temperature, humidity, vibration, and other relevant conditions.\n- Time and sequence: whether warming, drift, carryover, or other order effects occur.\n- Replication: enough observations to estimate dispersion with useful precision.\n\nRepeatedly recording an unchanged display is not equivalent to repeating a complete measurement. Likewise, successive readings from a slowly responding sensor may be correlated. Treating correlated readings as independent can make the apparent amount of information much greater than it really is.\n\nFor destructive tests, the same specimen cannot be measured repeatedly. Similar test items may instead be used, but variation among those items can contribute to the observed spread. Separating material heterogeneity from measurement variation may require an appropriate experimental design and statistical model."
    },
    {
      "heading": "Applications",
      "text": "Laboratory analysis\n\nIn analytical chemistry, repeatability is a standard characteristic in method validation and performance assessment. For chromatography, for example, repeatability may refer to repeated injections from one vial, independent preparations of one sample, or repeated execution of the complete analytical method. Each includes a different set of variability sources.\n\nA method can have excellent injection repeatability while showing substantially poorer repeatability when extraction, dilution, and sample preparation are included.\n\nManufacturing and inspection\n\nRepeatability helps characterize dimensional instruments, automated inspection systems, and production tests. Measurement-system studies distinguish variation obtained with the same operator and equipment from additional variation associated with changing operators or other conditions.\n\nIn robotics and positioning equipment, repeatability often describes the dispersion of positions reached after repeated commands to the same target. This differs from positioning accuracy, which concerns displacement from the intended position. A mechanism can repeatedly reach nearly the same incorrect location.\n\nBiological research\n\nBiological experiments often distinguish technical replicates from biological replicates. Technical replicates assess variation within a measurement or processing procedure; biological replicates represent independently sampled biological units. Repeatedly measuring one specimen does not establish how much independent specimens vary.\n\nIn quantitative biology, repeatability also has a specialized statistical use: the proportion of total observed variance associated with persistent differences among individuals or other grouping units. It is often expressed using an intraclass correlation coefficient. This variance-ratio meaning is related to consistency across repeated observations but is not identical to the metrological definition."
    },
    {
      "heading": "Relationship to measurement uncertainty",
      "text": "Repeatability commonly contributes to measurement uncertainty, but it is not a complete uncertainty evaluation.\n\nFor \\(n\\) independent observations with a stable common variance, the standard uncertainty associated with their mean can be estimated as\n\n\\[u_{\\mathrm{rep}}(\\bar{x})=\\frac{s_r}{\\sqrt{n}}.\\]\n\nThis reduction applies to the random component represented by the repeatability estimate. Averaging does not automatically reduce calibration uncertainty, persistent bias, uncertainty in reference values, or effects common to all observations.\n\nA full uncertainty evaluation may also include instrument resolution, corrections, environmental influences, sample preparation, and uncertainty associated with metrological traceability. Components already represented by the observed repeatability should not be counted a second time without justification."
    },
    {
      "heading": "Interpretation and limitations",
      "text": "Repeatability is conditional, rather than an intrinsic number attached permanently to an instrument or method. A reported value depends on the material, measurement range, operator, procedure, environment, and extent of the operation being repeated.\n\nVery high apparent repeatability can be misleading when an instrument rounds readings coarsely, filters signals strongly, or repeatedly measures an unrepresentative specimen. Identical displayed values do not demonstrate zero uncertainty.\n\nShort-term agreement also provides limited evidence about performance over months, across laboratories, or with different materials. Stable systematic errors can remain almost invisible in a repeatability study. Consequently, a useful report specifies the repeatability conditions, number of observations, statistical measure, measurement level or range, and whether repetition covered instrument readings alone or the complete procedure."
    }
  ],
  "links": [
    "metrology",
    "calibration",
    "reproducibility",
    "variance",
    "confidence-interval",
    "temperature",
    "analytical-chemistry",
    "chromatography",
    "measurement-uncertainty",
    "metrological-traceability"
  ],
  "sources": [],
  "license": "https://creativecommons.org/licenses/by/4.0/"
}