Cumulative hazard function Created 2026-09-24 Updated 2026-10-14
Suppose has an absolutely continuous probability distribution, with survival function . On an interval where , its cumulative hazard is
Here is the hazard function; its integral is dimensionless because the units of rate and time cancel. The identity follows by integrating the derivative of , so mere continuity of the survival function is not enough: the absolute continuity assumption supplies a probability density function. If reaches zero, interpret as . The normalization also matters; an initial probability mass at zero would require an initial term .
Accumulating hazard functions adds on the scale and multiplies survival factors on the scale. For and ,
For instance, a cumulative hazard increment of gives event probability , rather than a probability of . With per year, and . With the increasing hazard function per year when is expressed in years, , so gives . Its coefficient has units of inverse years squared.
Figure 1.
Constant and increasing hazard functions, their cumulative hazards, and their survival functions
. Details.
The constant hazard function produces an exponential distribution; the increasing example is a Weibull distribution. Their survival functions both decrease, although one hazard function stays constant and the other increases. At year ten both have and despite different earlier histories. The accumulation identity is described in Lu Tian's Stanford notes, Section 1.1.
For a discrete hazard at each possible event time , the exact counterpart is . Taking its negative logarithm sums , rather than . Therefore exponentiating a Nelson–Aalen estimator does not give exactly the Kaplan–Meier estimator; small-jump comparison of cumulative hazard estimators quantifies the difference.
For , the standard negative Weibull distribution has distribution function for and one for . This is the probability distribution of the negative of a standard positive Weibull distribution variable.
Normalize the probability density function using :
The corresponding Weibull distribution has survival function
Since the distribution is continuous, the choices and have the same probabilities. The conditional probability is consequently
For and , the supplied inequality, applied to , gives . Taking exponentials yields
This expresses the increasing hazard function : surviving to a later age shortens the remaining lifetime in this conditional distribution. At equality holds, as expected from the memoryless property of the exponential distribution.
Assume . Draw an independent component indicator with a Bernoulli distribution of success probability . On success, generate the normal distribution from part (i) and return its absolute value; this has the half-normal distribution probability density function on . On failure, generate the Weibull distribution from part (ii). Use fresh independent uniforms for the chosen component.
The law of total probability gives the output probability density function
as required. The factor two belongs inside the half-normal component; the probability of choosing that component is , not .
For , integrating the Weibull distribution probability density function gives
The inverse transform sampling method takes uniform on and solves . Hence
Since is also uniform, is an equivalent implementation. For any , the resulting event has probability , verifying its distribution.
For a continuous survival distribution, , and . Applying these identities gives the Weibull distribution
The survivor function has value one at zero; the hazard may diverge there when , which does not invalidate the distribution.
For common shape , the hazard ratio is , independent of time. Thus the two distributions lie in the same proportional hazards family. They also satisfy
Hence the time multiplier is and the hazard multiplier is , establishing the common accelerated life family as well. Here is reciprocal scale, not the ordinary time-scale parameter.
Put . With common Weibull distribution parameters and the hazard from part (a), substitution into the general log-likelihood gives
This is up to censoring-mechanism terms independent of . For and fixed , its maximizing rate is ; the shape still requires separate estimation if it is unknown.
A proportional hazards family has a common baseline hazard and positive constants such that . Thus its hazard ratios are constant in time, and its survivor functions satisfy . An accelerated life family instead has positive constants such that has the distribution of . Equivalently, : all time quantiles are multiplied by the same factor. These two properties are generally different.
Integrating the printed Weibull distribution density, or differentiating the following expression and checking its initial value, gives
Here is an inverse-time parameter: the coefficient of in the cumulative hazard is , not . This matters when comparing different parameterizations of a Weibull distribution.
With common shape , the hazard ratio is
It does not depend on , proving membership in one proportional hazards family. For the other property,
Consequently they also belong to one accelerated life family. A larger inverse-time parameter gives shorter survival times and larger hazards, consistently in both descriptions. This is the common-shape case of Weibull accelerated-life and proportional-hazards families.
For the stated scale-parameterisation of the Weibull distribution, its incubation survivor function is for , with . The continuous back-calculation of infection incidence equation becomes
Set before the infection process's start if one is specified. For the equal-width endpoint approximation, integrate the incubation density over each delay bin:
Thus the discrete Weibull equation is
For unequal intervals replace by . These are probabilities, not point evaluations of a density; no additional factor of is needed after bin integration.
For an exact failure half an hour after entry, use the conditional density, rather than the probability of an event at one exact time:
Again this is independent of by memorylessness of the exponential distribution. Whether successive observed lecture segments came from the same bulb does not affect the conditional likelihood: surviving segments of one bulb multiply into its survival exposure, and replacement starts another memoryless lifetime. Assume no unrecorded use relevant to the exposure calculation or, equivalently, condition on the bulb being alive at each observed lecture's start; unobserved ages themselves need not be known.
The observed working exposure is bulb-hours with one failure. Thus
Strict concavity of verifies the maximum. Immediate replacement would increase observed working exposure to 13 hours, since the replacement survives the remaining half hour, giving
For the alternative Weibull distribution and unknown entry age in a Weibull survival segment, keep its different inverse-scale convention: . Conditional survival for an additional duration is
and the conditional failure density after is
Both depend on the unknown age unless . The available segment data, with ages and bulb identity unrecorded, do not specify the usual Weibull likelihood needed to estimate . Merely collecting more of the same unknown-age records does not fix the missing age/design information. A larger dataset with documented ages or identities, or a justified model for the entry-age distribution, could permit estimation. If is imposed, the exponential calculation remains valid. The issue is loss of memorylessness and unspecified entry history, not a universal impossibility of learning from any larger bulb experiment.
Integrating the Weibull distribution density gives for . Inverse transform sampling therefore gives
Replacing by produces the same probability distribution. Both uniforms belong to with probability one. Here the conventional Weibull scale is , rather than .
Set . For fixed and large ,
This endpoint maximum likelihood for a singular location density calculation gives a Weibull distribution with shape . The estimator is a consistent estimator, and its error has order , faster than the error of . Its bias is asymptotic to and its mean squared error to , versus a positive constant times for the sample-mean estimator. Thus the minimum estimator is preferable for its much faster convergence, despite its positive finite-sample bias. This is a nonregular support-dependent model; ordinary regular-model efficiency bounds do not apply.
For independent and identically distributed random variables with distribution function , let be the sample maximum. In extreme value theory, belongs to the maximum domain of attraction of a nondegenerate distribution function when there exist and such that
at every continuity point of . Thus the normalized sample maxima have convergence in distribution to .
A positive measurable function has regular variation at infinity with index , written , if
The case defines a slowly varying function; equivalently a regularly varying function is with a slowly varying function.
Write for the survival function and for the right endpoint of a distribution. For , use the following standard extreme value theory normalization conventions:
They are respectively the Fréchet distribution, the negative Weibull distribution, and the Gumbel distribution for maxima. The negative Weibull distribution is supported to the left of its finite endpoint; it is not the usual positive Weibull distribution.
The necessary and sufficient maximum domain of attraction criteria are:
  • For the Fréchet distribution,
    Equivalently for every .
  • For the negative Weibull distribution,
    Equivalently as , for every .
  • For the Gumbel distribution,
    Here is a Gumbel auxiliary function, and can be finite or infinite. For each fixed , the shifted argument lies below eventually. This exponential tail-ratio condition, rather than regular variation with a fixed finite index, characterizes this case.
For the corresponding sufficient hazard function conditions, assume is absolutely continuous near , its survival function tends to zero there, and its probability density function is positive there. Write . The Von Mises conditions for extreme values give
The last condition additionally assumes that the reciprocal hazard function is continuously differentiable near the endpoint. It concerns the derivative of the reciprocal hazard function, not the derivative of the hazard function itself. Equivalently, when is differentiable, it is .
To see the connection, taking an integral of the hazard function gives
The first two limits make the logarithmic tail ratios over multiplicative or endpoint-distance intervals tend to or , giving the required regular variation. For the Gumbel distribution, set . The derivative condition makes uniformly on bounded -intervals, so
The endpoint assumptions ensure these shifts are admissible: at an infinite endpoint , and at a finite endpoint . These are sufficient smooth-tail conditions; the tail-ratio characterizations above are necessary and sufficient without differentiability assumptions. For a source using the same reciprocal-hazard convention, see Smith and Weissman's university lecture notes, equations (2.18)–(2.24).
For the Gamma distribution with integer shape and rate one, let . The base case is , and integration by parts gives
Induction proves
This survival function has the asymptotic form ; for the relation is exact. Put . For every fixed real , and . The proposed centering satisfies
so
Only the highest-degree term contributes to the limit, with the same reasoning covering . The Taylor expansion now gives
because . Therefore the Gumbel limit for gamma maxima is
Equivalently, has convergence in distribution to the standard Gumbel distribution, with scale .
The transformation is . Because has a unit-rate exponential distribution,
Thus has a Weibull distribution, with
Consequently is constant, proving proportional hazards.
Shape-two Weibull scale likelihood-ratio test Created 2026-10-31 Updated 2026-11-05
For independent Weibull distribution observations of known shape two, the likelihood ratio between any scale and the null scale increases with . Under the null, has a chi-squared distribution with degrees of freedom. The same upper-tail critical region is therefore most powerful for every larger scale, giving a uniformly most powerful size- test.
Squared Rayleigh radius 2026-10-18
Squaring a Rayleigh distribution variable produces an exponential distribution. For ,
where the exponential distribution parameter is a rate. Equivalently, has chi-squared distribution with two degrees of freedom (statistics). An amplitude and its square therefore have different probability density functions, even though one determines the other.
This makes the Rayleigh-normal scale mixture transparent: at unit scale, is exponential, allowing a random normal scale to produce a Laplace distribution. Another equivalent parametrization uses a Weibull distribution with shape two and scale . Conventions that call this latter scale the Rayleigh scale differ by a factor of ; check the probability density function before comparing numerical parameter values.[ref]
Weibull distributions with a common shape form both an accelerated life family and a proportional hazards family. Their hazard functions are , whose ratio is . Their time multiplier is . In the representation with and density for , the parameters are and , giving time multiplier and hazard multiplier .
Weibull distribution Created 2026-09-24 Updated 2026-11-05
The two-parameter Weibull distribution is a positive lifetime model with shape and scale . Its Probability density function and Survival function are
Its Cumulative distribution function is . The Hazard function is : it decreases for , is constant for , and increases for . This permits decreasing or increasing failure rates within one family.[ref]
The mean is and the variance is . The substitution reduces the power-moment integrals to gamma integrals. At the law is exponential with rate ; at its survival is .
Inverse transform sampling is explicit: for uniform on , take . Conditional survival after an already attained age is , which depends on unless . This is why an Unknown entry age in a Weibull survival segment affects the likelihood. The scale parameter here is ; inverse-scale parameterizations instead use its reciprocal.