{
  "id": "kaplan-meier-estimator",
  "title": "Kaplan–Meier Estimator",
  "lang": "en",
  "source_lang": "en",
  "summary": "The Kaplan–Meier estimator is a nonparametric estimator of a survival function that accommodates right-censored time-to-event observations.",
  "category": "mathematics",
  "subcategory": "probability-statistics",
  "type": "concept",
  "aliases": [
    "Product-limit estimator",
    "Kaplan–Meier survival estimator",
    "Kaplan–Meier estimate"
  ],
  "generated_at": "2026-11-05",
  "model": "pro/gpt-6.1-sol",
  "url": "https://aiwiki.page/en/kaplan-meier-estimator/",
  "translations": {
    "zh-Hans": {
      "url": "https://aiwiki.page/zh/kaplan-meier-estimator/",
      "json": "https://aiwiki.page/zh/kaplan-meier-estimator.json"
    }
  },
  "sections": [
    {
      "heading": "",
      "text": "The Kaplan–Meier estimator is a statistical method for estimating the probability that an event has not occurred by a specified time. It is a central tool in survival analysis, particularly when some observations are right-censored: their event times are unknown, but are known to exceed the time at which observation ended. The estimator requires no specified parametric distribution for event times and produces a step-shaped survival curve. Its applications include estimating patient survival, equipment reliability, and the duration of economic or social processes."
    },
    {
      "heading": "Statistical setting",
      "text": "Let \\(T\\) be a nonnegative random variable representing the time from a defined starting point to an event. Its survival function is\n\n\\[S(t)=P(T>t).\\]\n\nThus, “survival” means remaining free of the event under study; the event need not be death. For example, it may be a component failure or the termination of an employment spell.\n\nIf \\(F(t)=P(T\\leq t)\\) is the cumulative distribution function, then\n\n\\[S(t)=1-F(t).\\]\n\nWith right censoring, the available information for subject \\(i\\) is typically written as\n\n\\[X_i=\\min(T_i,C_i), \\qquad \\delta_i=\\mathbf{1}\\{T_i\\leq C_i\\},\\]\n\nwhere \\(C_i\\) is the censoring time and \\(\\delta_i\\) indicates whether the event was observed. A censored observation establishes that the subject remained event-free up to the censoring time; it does not establish what happened afterward.\n\nExamples of censoring include a study ending before a participant experiences the event or a functioning device being removed from observation. Whether censoring permits valid estimation depends on its relationship to the underlying event process."
    },
    {
      "heading": "Definition and calculation",
      "text": "Let\n\n\\[t_1<t_2<\\cdots<t_k\\]\n\nbe the distinct observed event times. At each \\(t_j\\), define:\n\n- \\(n_j\\): the number of subjects still under observation and event-free immediately before \\(t_j\\), called the risk set size;\n- \\(d_j\\): the number of observed events at \\(t_j\\).\n\nThe Kaplan–Meier estimator is\n\n\\[\\widehat S(t) = \\prod_{t_j\\leq t} \\left(1-\\frac{d_j}{n_j}\\right).\\]\n\nBefore the first event, the empty product equals \\(1\\). Each factor estimates the conditional probability of surviving an event time, given survival up to that point. Multiplying the factors estimates the probability of surviving through all preceding event times.\n\nThe resulting curve is right-continuous, remains constant between event times, and drops at observed events. Censoring does not directly cause a drop: it reduces the risk set for later events. Plots often mark censoring times with short ticks.\n\nSeveral events recorded at the same time are handled together through \\(d_j\\). When events and censoring share a recorded time, the usual convention includes those censored at that time in the risk set for the events, then removes them afterward. This convention matters when time measurements are coarse.\n\nExample\n\nSuppose five devices are observed:\n\n- failures occur at times \\(2\\), \\(4\\), and \\(6\\);\n- one device is censored at time \\(3\\);\n- another is censored at time \\(7\\).\n\nThe calculation is:\n\nEvent time \\(t_j\\) | At risk \\(n_j\\) | Events \\(d_j\\) | Estimated survival after the event\n2 | 5 | 1 | \\(4/5=0.80\\)\n4 | 3 | 1 | \\((4/5)(2/3)=8/15\\)\n6 | 2 | 1 | \\((8/15)(1/2)=4/15\\)\n\nThe censoring at time \\(3\\) explains why only three devices remain at risk immediately before time \\(4\\). At time \\(6\\), the estimated survival probability falls to approximately \\(0.267\\).\n\nIf no observations are censored, the product telescopes and gives\n\n\\[\\widehat S(t) = \\frac{\\text{number of observed event times greater than }t} {\\text{total number of subjects}}.\\]\n\nIt therefore reduces to the empirical survival function for completely observed data."
    },
    {
      "heading": "Assumptions and statistical interpretation",
      "text": "The standard formulation assumes independent subjects and a censoring mechanism that does not selectively remove subjects according to their unobserved event times. A common sufficient condition is independence between \\(T\\) and \\(C\\).\n\nMore generally, censoring must be noninformative for the survival distribution being estimated: among subjects still at risk, those censored should have the same subsequent event prospects as comparable subjects who remain observed. This is not guaranteed merely because censoring is recorded correctly.\n\nFor example, if devices showing signs of imminent failure are preferentially withdrawn, the observed failures need not represent the experience of all devices. The ordinary Kaplan–Meier curve may then overestimate survival. Independence of event and censoring times only after conditioning on covariates does not generally justify a pooled, unadjusted curve; stratification, modeling, or weighting may be needed.\n\nUnder independent censoring and appropriate observation conditions, the estimator is a consistent estimator of \\(S(t)\\) on the time range supported by follow-up. Its nonparametric character means that it does not assume a particular distributional family, such as an exponential distribution. It does not mean that the analysis is assumption-free.\n\nThe estimator also has a nonparametric maximum-likelihood interpretation. At an event time, the estimated conditional event probability is \\(d_j/n_j\\), and multiplying the complementary probabilities gives the product-limit formula. If the largest observed time is censored, the remaining tail probability cannot be located from the observed data alone."
    },
    {
      "heading": "Uncertainty and confidence intervals",
      "text": "The most widely used estimate of the estimator’s variance is Greenwood’s formula:\n\n\\[\\widehat{\\operatorname{Var}}\\{\\widehat S(t)\\} = \\widehat S(t)^2 \\sum_{t_j\\leq t} \\frac{d_j}{n_j(n_j-d_j)}.\\]\n\nIts square root is the estimated standard error. The displayed formula requires care at boundary cases, particularly when all remaining subjects experience the event and \\(n_j=d_j\\).\n\nA simple normal-approximation confidence interval is\n\n\\[\\widehat S(t) \\pm z_{1-\\alpha/2} \\sqrt{\\widehat{\\operatorname{Var}}\\{\\widehat S(t)\\}},\\]\n\nwhere \\(z_{1-\\alpha/2}\\) is a standard normal quantile. Such intervals may extend outside \\([0,1]\\) and can perform poorly near the boundaries. Transformed intervals, often based on the complementary log–log transformation\n\n\\[g(S)=\\log[-\\log S],\\]\n\nare commonly used to keep interval endpoints within the allowable range.\n\nA pointwise interval describes uncertainty at a particular time. It is not a simultaneous confidence band for the entire curve. Uncertainty generally becomes greater late in follow-up as the risk set shrinks, so a curve’s tail must be interpreted alongside the number still at risk."
    },
    {
      "heading": "Derived summaries",
      "text": "Median survival\n\nThe estimated median survival time is conventionally defined as\n\n\\[\\widehat m = \\inf\\{t:\\widehat S(t)\\leq 0.5\\}.\\]\n\nIt is the first time the curve reaches or falls below one half. If the observed curve never reaches that level, the median is not estimable from the available follow-up; this does not imply that the population median is infinite.\n\nRestricted mean survival time\n\nThe restricted mean survival time up to a specified horizon \\(\\tau\\) is\n\n\\[\\operatorname{RMST}(\\tau) = \\int_0^\\tau S(t)\\,dt.\\]\n\nReplacing \\(S\\) with \\(\\widehat S\\) estimates this quantity as the area under the Kaplan–Meier curve through \\(\\tau\\). It represents the expected event-free time accumulated within that horizon.\n\nThe unrestricted mean requires information about the entire tail. When follow-up ends with a positive estimated survival probability, it is generally not identifiable without additional assumptions."
    },
    {
      "heading": "Related methods and limitations",
      "text": "Kaplan–Meier curves can be constructed separately for groups. The log-rank test is a standard method for comparing their survival experience, but it is a separate inferential procedure rather than part of the estimator. Neither a visible separation between curves nor a statistically significant comparison establishes a causal effect: group differences may reflect confounding or selection.\n\nThe Cox proportional hazards model extends time-to-event analysis to regression with covariates. Unlike a simple Kaplan–Meier curve, it explicitly models the relationship between covariates and event rates.\n\nImportant limits of the basic estimator include:\n\n- Competing risks. When another event prevents the event of interest, treating that competing event as censored and reporting \\(1-\\widehat S(t)\\) does not generally estimate the observed probability of the event of interest. That probability is described by a cumulative incidence function, commonly estimated using the Aalen–Johansen estimator.\n- Interval censoring. If an event is known only to have occurred between two examinations, its exact time is unavailable. Assigning it to an examination time and applying ordinary Kaplan–Meier estimation does not properly represent that observation structure.\n- Delayed entry. Subjects who enter observation after the chosen time origin require risk sets that account for their entry times. This is often called left truncation and differs from censoring.\n- Dependent observations. Repeated events or observations clustered within families, institutions, or devices may require methods that account for dependence.\n- Unsupported extrapolation. A horizontal final segment does not establish that event risk has disappeared or that the remaining subjects will never experience the event.\n\nThe estimator concerns survival probabilities, not instantaneous event rates. Although \\(d_j/n_j\\) represents an estimated conditional event probability at an observed event time, the Kaplan–Meier curve is not itself a smooth hazard-rate estimate."
    },
    {
      "heading": "Historical development",
      "text": "Edward L. Kaplan and Paul Meier introduced the estimator in their 1958 paper, Nonparametric Estimation from Incomplete Observations. Its name “product-limit” reflects its construction as a product of conditional survival estimates. It became a standard method for individually recorded event times, complementing grouped life-table methods that summarize experience over time intervals."
    }
  ],
  "links": [
    "statistics",
    "random-variable",
    "cumulative-distribution-function",
    "conditional-probability",
    "statistical-independence",
    "consistent-estimator",
    "exponential-distribution",
    "maximum-likelihood-estimation",
    "variance",
    "standard-error",
    "confidence-interval",
    "median",
    "expected-value",
    "confounding",
    "cox-proportional-hazards-model",
    "life-table"
  ],
  "sources": [],
  "license": "https://creativecommons.org/licenses/by/4.0/"
}