1 Definition and basic form

The Weibull distribution is a continuous probability distribution used to describe the time until an event occurs. Typical examples include failure of a component, onset of a disease, or the length of a survival time. It is widely used because it can model several different shapes of failure behavior with a simple set of parameters.

The distribution is defined on nonnegative values in its simplest form, although a location parameter can shift it along the time axis. Its flexibility makes it a standard tool in reliability engineering and survival analysis.

1.1 Probability density function

For a two-parameter Weibull distribution with shape parameter \(k>0\) and scale parameter \(\lambda>0\), the probability density function is

\[ f(x)=\frac{k}{\lambda}\left(\frac{x}{\lambda}\right)^{k-1}e^{-(x/\lambda)^k}, \quad x\ge 0. \]

This density is right-skewed for many parameter values, but its exact form depends strongly on the shape parameter. When \(k<1\), the density is highest near zero and decreases thereafter; when \(k>1\), it rises from zero, reaches a peak, and then declines.

1.2 Cumulative distribution function

The cumulative distribution function gives the probability that the event has occurred by time \(x\):

\[ F(x)=1-e^{-(x/\lambda)^k}, \quad x\ge 0. \]

This expression is simple and analytically convenient. It is often used to compute percentiles and to fit failure-time data.

1.3 Survival function

The survival function is the complement of the cumulative distribution function:

\[ S(x)=P(X>x)=e^{-(x/\lambda)^k}. \]

In applications, this represents the probability that a component or subject has not yet failed or experienced the event by time \(x\). The survival function is central to reliability analysis because it directly describes remaining life.

1.4 Hazard function

The hazard function, or failure rate, is

\[ h(x)=\frac{f(x)}{S(x)}=\frac{k}{\lambda}\left(\frac{x}{\lambda}\right)^{k-1}. \]

This function can be increasing, constant, or decreasing depending on the shape parameter. That property is one of the main reasons the Weibull distribution is so widely used.

1.5 Parameterization

Several parameterizations exist in the literature. The most common uses a shape parameter, a scale parameter, and sometimes a location parameter. Different fields may exchange notation, but the underlying distribution remains the same.

1.5.1 Shape parameter

The shape parameter controls the form of the distribution and the behavior of the hazard rate. It is often denoted by \(k\) or \(\beta\). Small values produce a steeply declining failure rate, while larger values create an increasing failure rate.

1.5.2 Scale parameter

The scale parameter, often written as \(\lambda\) or \(\eta\), stretches or compresses the distribution along the horizontal axis. Larger values spread the distribution over a longer time range. In practical terms, it sets the characteristic life of the process.

1.5.3 Location parameter

A three-parameter form includes a location parameter, usually denoted by \(\gamma\), which shifts the distribution to the right. In that case the variable is modeled as \(x-\gamma\) rather than \(x\). This form is used when failures cannot occur before a known threshold time.

The Weibull distribution is connected to several familiar distributions. Some are obtained as limiting or special cases, while others arise through transformations. These relationships make the Weibull distribution useful in theory as well as in applications.

2.1 Exponential distribution

When the shape parameter equals 1, the Weibull distribution becomes the exponential distribution. The hazard rate is then constant, which corresponds to a memoryless failure process. This special case is important in reliability and queueing models.

2.2 Rayleigh distribution

When the shape parameter equals 2 and the scale is chosen appropriately, the Weibull distribution reduces to the Rayleigh distribution. The Rayleigh form appears in modeling magnitudes of two-dimensional random vectors and in some signal-processing contexts.

2.3 Extreme value distribution

The Weibull distribution is related to extreme value theory through transformations of minima and maxima. In particular, one form of the Weibull appears as the distribution of extreme values under certain normalization schemes. This connection is one reason the distribution is prominent in statistical theory.

2.4 Generalized Weibull family

Broader families extend the basic Weibull form by adding shape parameters or mixture structures. These generalizations increase flexibility but often reduce mathematical simplicity. They are used when real data show behavior that the standard Weibull cannot capture well.

3 Statistical properties

The Weibull distribution has many tractable properties. Several of them can be expressed using the gamma function, which simplifies theoretical work and numerical computation. The exact form of these properties depends on the chosen parameterization.

3.1 Moments

Raw moments can be written in closed form. For a two-parameter Weibull random variable \(X\),

\[ E[X^r]=\lambda^r\Gamma\!\left(1+\frac{r}{k}\right), \]

provided \(r>-k\). This formula is useful for deriving measures such as the mean, variance, and higher-order characteristics.

3.2 Mean and variance

The mean is

\[ E[X]=\lambda\Gamma\!\left(1+\frac{1}{k}\right). \]

The variance is

\[ \lambda^2\left[\Gamma\!\left(1+\frac{2}{k}\right)-\Gamma^2\!\left(1+\frac{1}{k}\right)\right]. \]

These expressions show how both the shape and scale parameters influence the center and spread of the distribution.

3.3 Median and mode

The median is obtained by solving \(F(x)=1/2\), giving

\[ m=\lambda(\ln 2)^{1/k}. \]

The mode exists only for \(k>1\) and is

\[ x_{\text{mode}}=\lambda\left(\frac{k-1}{k}\right)^{1/k}. \]

When \(k\le 1\), the distribution is monotone and its highest value occurs at the boundary near zero.

3.4 Skewness and kurtosis

Skewness and kurtosis can also be calculated from the moments. Their values depend strongly on the shape parameter. As \(k\) increases, the distribution becomes less skewed and more concentrated, while small values of \(k\) produce heavier right tails and stronger asymmetry.

3.5 Entropy

The differential entropy of the Weibull distribution has a closed-form expression involving the Euler–Mascheroni constant and the logarithm of the scale parameter. Entropy summarizes the overall spread or uncertainty of the distribution. It is often used in information-theoretic studies and model comparison.

4 Parameter estimation

Estimating Weibull parameters from data is a major practical task. Several approaches are available, and the best choice depends on sample size, censoring, and the intended use of the model. Reliability and survival settings often require methods that handle incomplete observations.

4.1 Maximum likelihood estimation

Maximum likelihood estimation is the most common inferential method. It chooses parameter values that maximize the probability of the observed data. The resulting equations are usually solved numerically, especially when censored data are present.

4.2 Method of moments

The method of moments matches sample moments with theoretical moments. It is conceptually simple and can provide initial parameter estimates. However, it is often less efficient than maximum likelihood and may perform poorly for heavily skewed data.

4.3 Linear regression methods

Linearized estimation methods transform the Weibull relationship into a straight line on special plotting paper or under logarithmic transformations. These approaches have long been used in engineering practice because they are visually intuitive. They are generally less precise than full likelihood-based methods but remain useful for quick diagnostics.

4.4 Bayesian estimation

Bayesian estimation combines data with prior information about the parameters. It is particularly helpful when samples are small or prior engineering knowledge is available. Posterior distributions are usually computed with numerical methods such as Markov chain Monte Carlo.

5 Applications

The Weibull distribution appears in many applied fields because it captures a wide range of time-to-event patterns. Its parameters have practical interpretations, which makes the model easier to explain and communicate than many alternatives.

5.1 Reliability engineering

In reliability engineering, the distribution is used to describe component lifetimes and system durability. Engineers use it to estimate failure probabilities, compare product designs, and plan maintenance schedules. Its hazard function is especially valuable for distinguishing wear-out, random failure, and early-life failure behavior.

5.2 Survival analysis

In survival analysis, the Weibull model is used for time-to-event data involving subjects who may be censored. It can represent survival times in medical, biological, or social settings. Because the hazard can increase or decrease, it is more flexible than the exponential model.

5.3 Failure time modeling

The distribution serves as a general model for failure times in settings ranging from electronics to mechanical parts. It is often employed when the time to failure is influenced by aging, stress, or material degradation. The model can also support accelerated life testing.

5.4 Wind speed modeling

Weibull distributions are commonly used to approximate wind speed data. In this context, the shape and scale parameters describe how frequently different wind speeds occur. The model is important in renewable energy studies, particularly for estimating wind resource availability.

5.5 Materials and fatigue analysis

In materials science, the distribution is applied to fracture, fatigue, and strength variability. It can characterize the distribution of weak points, crack initiation times, or time to material failure. Engineers often use it when analyzing lifetime under cyclic loading.

6 Interpretation of parameters

The parameters of the Weibull distribution have direct practical meaning. Their effects are often easier to interpret than those of more abstract statistical models, which contributes to the distribution’s popularity in applied work.

6.1 Failure-rate behavior

The shape parameter determines the direction of the failure rate over time. Values below 1 indicate decreasing failure rate, often associated with early defects. A value of 1 gives a constant rate, while values above 1 imply increasing failure rate, which is typical of wear-out processes.

6.2 Shape effects on distribution form

As the shape parameter changes, the density can become highly concentrated near zero, roughly symmetric over a limited range, or strongly peaked. Small values produce a long tail, while larger values tighten the distribution around its central region. This adaptability allows the same family to represent different physical processes.

6.3 Scale effects on spread

The scale parameter changes the typical time scale without altering the general qualitative shape. Increasing it shifts mass toward larger values and widens the spread. In practical terms, it often corresponds to the characteristic lifetime of the system under study.

7 Inference and diagnostics

Model fitting is usually followed by diagnostic checks. These help determine whether the Weibull distribution is a reasonable description of the data and whether the estimated parameters are stable enough for prediction or decision-making.

7.1 Goodness-of-fit tests

Goodness-of-fit tests assess whether the observed data are consistent with a Weibull model. Common tools include likelihood-based tests, Anderson-Darling methods, and comparisons with alternative distributions. No single test is decisive, so results are typically interpreted alongside graphical checks.

7.2 Probability plots

Weibull probability plots transform the data so that a correct model appears approximately linear. These plots are useful for visual assessment of fit, identification of outliers, and comparison of subsamples. They have long been a standard engineering diagnostic.

7.3 Residual analysis

Residual analysis examines how far observed values deviate from fitted values under the model. In censored or survival settings, residuals can reveal lack of fit, unusual observations, or time-dependent effects not captured by the Weibull form. The method complements formal tests and probability plots.

8 Variants and generalizations

Several variants of the Weibull distribution are used in practice. These extensions allow for more complex behavior, such as multiple failure mechanisms or additional flexibility in the tail. They are especially useful when real data depart from the standard shape.

8.1 Three-parameter Weibull distribution

The three-parameter Weibull distribution adds a location shift. This version is useful when there is a minimum possible failure time or a delayed onset of risk. It is more flexible than the two-parameter form, but it can be harder to estimate reliably.

8.2 Two-parameter Weibull distribution

The two-parameter Weibull distribution is the standard version with shape and scale only. It is the most widely used form because of its simplicity and interpretability. In many applications it is sufficient to describe the observed data well.

8.3 Weibull mixture models

Mixture models combine two or more Weibull components to represent heterogeneous populations or multiple failure modes. They can capture multimodal or more complex lifetime patterns. Such models are often used when a single Weibull distribution is too restrictive.

8.4 Generalized gamma and competing models

The generalized gamma distribution and other competing lifetime models provide alternative ways to model skewed time-to-event data. These families may fit better when hazard rates or tail behavior differ from the Weibull pattern. Model comparison is typically based on likelihood, diagnostics, and interpretability.

9 History

The Weibull distribution is named after Waloddi Weibull, who promoted and analyzed it in the early twentieth century. Its later adoption across engineering and statistics reflected both practical usefulness and mathematical tractability.

9.1 Waloddi Weibull

Waloddi Weibull was a Swedish engineer and scientist associated with the distribution that bears his name. He used it to model the strength of materials and the breaking of solids. Although the form had earlier precedents, his work helped establish the distribution in applied analysis.

9.2 Early adoption in engineering

The distribution was quickly embraced in engineering because it matched observed lifetime data and supported design decisions. Its probability plots and parameter interpretations made it accessible to practitioners. Over time it became a standard tool in reliability testing and quality control.

9.3 Development in statistical theory

As statistical theory advanced, the Weibull distribution gained a firm place in probability and inference. Researchers studied its moments, estimation methods, censored-data behavior, and relationships to broader families. Today it remains a core example in both theory and application.