Inverting gives the quantile function
Consequently, inverse transform sampling uses with uniform on . The quartiles are , and .[ref]
The first and third quartiles are about and , so the central half of the probability spans about of the unit interval. Under a uniform distribution, its width would be .
The symmetric probability distribution gives . The angle representation gives a short calculation of the second moment:
Use to evaluate the integral. A uniform distribution on has variance ; the larger variance quantifies the greater spread toward the ends.
More generally, for a nonnegative integer , the beta function yields
For this is one; for it reproduces the values above. The central binomial coefficient thus appears in an exact moment calculation for a continuous probability distribution.
Exponential-logarithmic distribution Created 2026-10-25 Updated 2026-10-31
For , and , this probability density function has cumulative distribution function . Differentiating gives the density, while the endpoint limits are zero and one. Its quantile function is for , enabling inverse transform sampling.
Expanding the denominator as a geometric series gives
Termwise integration, justified by the monotone convergence theorem, and the gamma function integral yield raw moments . Thus the polylogarithm expresses its mean, variance and higher moments. Here is the complement of the series parameter ; parameter conventions must be checked when comparing formulas.
Glivenko-Cantelli theorem Created 2026-09-24 Updated 2026-11-05
For independent and identically distributed real observations with cumulative distribution function , the empirical distribution function
satisfies the Glivenko-Cantelli theorem
The population law may have atoms; continuity of is not required. The conclusion is uniform over every threshold, strengthening the separate strong laws for fixed thresholds.[ref]
Here is a proof using inverse transform sampling. First take iid uniforms and their empirical CDF . At the finitely many grid points , the strong law of large numbers gives simultaneously. Monotonicity bounds the discrepancy between grid points by the largest grid discrepancy plus . Taking countably many proves uniform convergence on .
For general , construct using the quantile function. Then exactly when , so and the general supremum is bounded by the uniform one. This also handles jumps of . The theorem does not apply merely because an empirical CDF was formed: for repeated copies of the same nondegenerate random observation, the empirical CDF does not approach its population law in this way.
Gumbel distribution Created 2026-10-05 Updated 2026-10-14
The standard Gumbel distribution for maxima has distribution function for every real . Its probability density function is the derivative of its distribution function:
This article uses the maxima convention throughout. Reflecting a standard Gumbel distribution random variable to gives the minima convention, with distribution function . The two conventions have opposite directions of asymmetry, so their formulas must not be interchanged. [ref].
For a location and positive scale , let . This location-scale family has cumulative distribution function and probability density function
Solving gives the standard quantile function , , and hence the standard median
The median of is therefore . Location is not the median: at the cumulative distribution function is .
To derive the Gumbel distribution as an extreme-value limit, take the maximum of independent exponential random variables with rate and set
The exact distribution function is
The finite- lower boundary is required because . For each fixed real , eventually , and the natural logarithm of the first branch satisfies
Exponentiating proves convergence in distribution to the standard Gumbel distribution. It is the familiar limit with : a simple calculation reveals a whole limiting shape. The name honors Emil Gumbel; Siegrist gives this exponential-maxima derivation and the standard Gumbel distribution formulas. [ref].
Figure 1.
Centered exponential sample maxima for n equal to 1, 5, 20 and 100 approach the standard Gumbel cumulative distribution function
. Details.
For the toy delay model , , the limiting Gumbel distribution approximation to has and . Its predicted median is
compared with the exact median . The approximation is about seconds lower in this example. It is not an exact finite-sample probability distribution: the approximating Gumbel distribution assigns positive probability to negative delays, while the exact sample maximum cannot be negative.
The maximum domain of attraction of the Gumbel distribution is characterized by an exponential tail ratio on the scale of a Gumbel auxiliary function. This allows many original probability distributions to share the limiting shape, but it does not make the centering appropriate to all of them.
Inverse transform sampling Created 2026-09-24 Updated 2026-10-14
If has a uniform distribution on and is the quantile function of a cumulative distribution function , then has the desired probability distribution. Indeed, right continuity gives
so . This short argument works for both jumps and smooth cumulative distribution functions. MIT's continuous-probability notes introduce the inverse-transformation construction. [ref].
For the three-value example, send to , to , and to . These intervals have exactly the desired widths , , . For the density example, output : for , . Independent uniform draws give independent random variables after this deterministic transformation. This direction of construction works even when has jumps; applying the probability integral transform in the other direction requires its own continuity hypotheses.
Left continuity of the quantile function Created 2026-10-06 Updated 2026-11-05
For , , the quantile function is a monotone function and is left-continuous. The distribution function is right-continuous, giving exactly when . If , a point between them has and contradicts this equivalence for some . A jump just to the right of is possible when is flat on an interval.
Log-logistic distribution Created 2026-10-31 Updated 2026-11-05
For a scale and shape , the log-logistic distribution has cumulative distribution function and probability density function
It has no point mass at zero. Its quantile function is for , so its median is . The log odds relation explains both its name and the log-logistic probability plot.[ref]
Its survival function is , with power tail proportional to . A positive moment is finite exactly when : near infinity its defining integral behaves like . For example, gives median and . Writing and is an alternative parameterization requiring .
Let be an integer and let be independent and identically distributed random variables with an exponential distribution of rate . Their sample maximum is nonnegative. Substituting the exponential distribution's cumulative distribution function into the distribution functions of independent minima and maxima gives the exact law
For example, at each observation is below the threshold with probability , but all observations are below it with probability . Thus an individual observation's median becomes an increasingly unlikely upper bound for the whole batch. The general order statistic density in Siegrist's treatment, with , also gives this law on integration. [ref].
Solving gives the exact quantile function, for ,
In a toy model with , the median maximum is seconds for , seconds for , and seconds for . All three use the exact distribution function; the increase comes from taking a maximum over more observations. The case connects to the existing maximum of two exponentials as a scaled sum.
Figure 1.
Exact cumulative distribution functions of exponential sample maxima, with rate one per second and median markers
. Details.
The Gumbel distribution explains the logarithmic shift visible here: after subtracting and measuring deviations in units of , the distribution functions approach one fixed shape. The next derivation makes that convergence in distribution explicit.
Median Created 2026-09-24 Updated 2026-11-05
A median of a real random variable is a number with at least half the probability at or below it and at least half at or above it:
Equivalently, its cumulative distribution function satisfies . The non-strict inequalities matter for atoms: if almost surely, zero is its median despite having no probability strictly below or strictly above it.
A median need not be unique. If zero and one each have probability , every point in is a median. The generalized inverse selects the lower median. For a continuous strictly increasing cumulative distribution function, the unique median solves .
For a finite data set, the sample median measures the middle rather than the arithmetic average. The sample has median three and mean 22; changing 100 to a still larger value leaves its median unchanged. With finite first absolute moment, medians are exactly the minimizers of expected absolute loss, giving a decision-theoretic reason for their usefulness.
Monotone rearrangement Created 2026-10-05 Updated 2026-10-06
For an atomless measure on with cumulative distribution function , the monotone transport to a target with quantile function is , defined -almost everywhere. It minimizes well-defined costs for convex continuous . With atoms in the source, the common-quantile transport plan remains available but need not be induced by a map.
Monotonic function 2026-10-14
A real-valued function , with domain of a function in the usual order of the real numbers, is a monotonic function if it is either a nondecreasing function throughout or a nonincreasing function throughout . The same direction must work for every pair in . Being a strictly increasing function on one part and a strictly decreasing function on another does not make the whole function monotone: for example, on the real line changes direction at zero. A constant function is both a nondecreasing function and a nonincreasing function.
The word "increasing" is ambiguous across textbooks: some allow equal outputs and some require unequal outputs. Use nondecreasing function for the weak inequality and strictly increasing function for the strict inequality; use nonincreasing function and strictly decreasing function in the other direction. Neither continuity nor differentiability is part of the definition of a monotonic function.
Strict monotonicity gives injectivity, so an equation has at most one solution and an inverse function exists on the image of a function. Weak monotonicity permits repeated outputs and jumps. In probability theory, a cumulative distribution function is a nondecreasing function; its quantile function uses a generalized inverse to handle those flat parts and jumps. Monotonicity alone does not guarantee continuity, even for a strictly increasing function.
Figure 1.
Strict and weak monotonicity: cubic functions, flat portions, and a step
. The strictly increasing function and strictly decreasing function never repeat an output. The nondecreasing function has flat intervals, and the step function for , for is a discontinuous cumulative distribution function. The open and closed circles specify its value at the jump; no vertical segment belongs to its graph. Details.
Nondecreasing function Created 2026-10-06 Updated 2026-10-14
A real-valued function is a nondecreasing function if every pair in its domain of a function satisfies . Flat intervals are allowed, unlike for a strictly increasing function. For example,
is a continuous nondecreasing function, but it repeats its output on each flat interval.
A nondecreasing function can also be discontinuous. The step function for and for has a jump discontinuity at zero without ever decreasing. It is the cumulative distribution function of a random variable that equals zero with probability one; quantile functions handle such jumps through a generalized inverse. Nondecreasing allocation functions permit critical-value payments in single-parameter mechanisms.
Nonregular identification at a flat mean map occurs when different parameter values give different probability distributions, yet the first derivative of the observable mean vanishes at a parameter of interest. Identifiability is a property of the entire parameter-to-distribution map; a first-order sensitivity calculation can miss it.
Consider an original statistical model with independent observations
The cubic mean map is a strictly increasing function, so equality of the normal distributions implies equality of . The model is globally identifiable, including at . Nevertheless, the single-observation score function and Fisher information are
Thus does not imply nonidentifiability. The rank is zero at the origin and one at every nonzero nearby point. This is precisely why a regular-point assumption matters: Rothenberg's local identification criterion assumes constant information-matrix rank in a neighborhood.[ref]
Maximum likelihood estimation gives , using the real cube root. At the true value zero, write with a standard normal random variable . Then
This is an exact sampling distribution identity for every , not merely a convergence in distribution statement. The estimation error has scale and a nonnormal shape. At a fixed nonzero , the usual delta method instead gives asymptotic variance ; that approximation cannot be extended uniformly through zero.
An exact confidence interval through a cubic mean map follows by applying the strictly increasing function to both endpoints of the ordinary known-variance confidence interval for the mean:
Here is the quantile function of the standard normal distribution. The transformed interval has exact coverage for every . When and the observed sample mean is zero, the 95% interval is approximately . Multiplying by 64 halves its width at that observed mean. A first-order interval based on dividing by the zero mean-map derivative would fail instead.
Figure 1.
Cubic mean map and exact confidence-interval scaling at an observed sample mean of zero; original normal-observation model
. Details.
Normal Q-Q plot 2026-10-14
A normal Q-Q plot compares order statistics with standard normal quantiles . For a normal distribution with , the population quantile function is , so the diagnostic sample relation is
The appropriate reference is a line with an intercept and positive slope, not necessarily the identity line. A line can be fitted through two central sample quantiles; R's qqline uses the quantile levels and by default. Neither that line nor its fitting rule is part of the definition of a normal Q-Q plot. [ref]
To isolate curvature, retain the same 19 plotting positions and construct two more ordered sequences:
The are continuous uniform distribution quantiles on , while the apply the hyperbolic sine to the reference quantiles. Both examples are symmetric about zero. At ranks , is approximately and is approximately . Again, these are constructed values rather than random draws or real data.
Figure 1.
Normal Q-Q curvature distinguishes compressed and stretched outer quantiles
. Left: the constructed uniform distribution quantiles bend inward relative to the central tangent . Right: the constructed values bend outward relative to the central tangent . Both panels use 19 standard normal quantiles at . Dashed lines are analytic central tangents, not fitted lines or uncertainty bands; only the chosen finite range is illustrated. Details.
Read these shapes relative to their central slope. The uniform distribution example has , with derivative ; at the upper endpoint the tangent predicts , while . Its lower end lies above the tangent and its upper end below it: the outer quantiles are compressed. The hyperbolic sine example has ; its tangent predicts at the upper endpoint, while . Its lower end lies below the tangent and its upper end above it: the outer quantiles are stretched. Unequal departures at the two ends can suggest skewness. The direction of these descriptions assumes that reference quantiles are horizontal and observed values vertical. The graphs illustrate finite-range shape differences; a sample pattern cannot establish an asymptotic tail probability law or membership in a heavy-tailed distribution class.
A nearly straight normal Q-Q plot is a diagnostic, not a proof of a normal distribution or a formal hypothesis test. Random statistical samples fluctuate, particularly at extreme order statistics, and apparent curvature depends on sample size, plotting positions and the reference line. A calibrated hypothesis test or uncertainty band requires its own assumptions and must account for parameters estimated from the same observations. Sorting also discards the original observation order: rearranging a time series leaves its Q-Q plot unchanged, so the plot cannot establish independence.
For a normal linear model, R's diagnostic normal Q-Q plot uses standardized regression residuals, also called internally studentized residuals. [ref] Even with independent errors, the regression residuals have covariance matrix , where is the hat matrix; fitting parameters generally creates dependencies. Dividing by a shared estimated standard deviation creates further dependence, and the distribution of an internally studentized Gaussian residual is not exactly the standard normal distribution. The normal Q-Q plot therefore helps assess error shape alongside other diagnostics rather than certifying independent normal random variables.
Pareto distribution Created 2026-10-07 Updated 2026-10-18
The Pareto distribution describes a positive random variable whose tail probability falls as a power law. In the shape–minimum convention used here, has shape parameter and minimum , with probability support . Its probability density function is
The minimum has the same units as ; the shape parameter is dimensionless. At fixed , changing gives a scale family, since . This convention is also used in uniform-Pareto conjugacy.[ref]
The survival function and cumulative distribution function are
For , solving gives the quantile function
Its right-hand limit at is ; it tends to infinity as .[ref] There is no probability mass at : changing a probability density function's value at that single point changes no probabilities.
Choose and as an original theoretical example. Then , , the median is , and the 0.99 quantile is . These chosen values and all figures below describe mathematical probability distributions, not actual wealth observations.
Figure 1.
Pareto survival curves: smaller alpha gives a slower power-law tail
. Details.
On this log-log plot, write to obtain . The slope is , whereas the probability density function has power exponent . Multiplying a threshold by ten divides its tail probability by . A larger therefore concentrates the Pareto distribution nearer ; a smaller makes extreme values more plausible. This exact scaling is useful when modelling large positive losses or sizes over several orders of magnitude.
The Pareto distribution is a heavy-tailed distribution: for every , the exponential moment diverges because eventually grows without bound. In contrast, the Laplace distribution has exponentially decreasing tails and some finite positive exponential moments. The Student's t-distribution also has power tails, but on both sides of the real line, with a smooth central region rather than a positive minimum.
Let be the distribution measure of . The measurable change-of-variables rule for expectations gives
For the second equality, let be uniform on and . The quantile function is measurable and satisfies
This equivalence uses monotonicity and right-continuity of . Consequently : has the same distribution as . In particular , so the signed integral is legitimate. Hence the quantile expectation formula is
No continuity or strict increase of the distribution function is required.
For a continuous strictly increasing cumulative distribution function and a uniform distribution on , monotonicity gives
Thus each transformed draw has distribution function . This is inverse transform sampling. The conclusion extends to discontinuous or non-strictly increasing by its quantile function : right continuity gives exactly when . Independent uniforms produce independent transformed draws, but marginal uniformity alone does not establish independence.
If “uniformly distributed sequence” instead means deterministic equidistribution, the empirical fraction of is the empirical fraction of and tends to . That is an empirical distribution statement, not a claim that the deterministic sequence consists of independent random variables.
Use the quantile function for , so a strictly increasing CDF is not required. For any real , the quantile event agrees, up to endpoint null sets, with . Hence
Thus has distribution function and probability density function . This proves inverse transform sampling, including a probability density function with intervals on which the CDF is constant.
For an invertible continuous cumulative distribution function, monotonicity gives
Thus each transformed observation has the desired distribution. More generally use the quantile function ; the same event identity proves inverse transform sampling even when there are atoms. Independent uniform inputs give independent transformed samples; mere uniform marginal distributions do not imply independence between different samples.
Group the two terms containing to obtain the mixture distribution
The first component is a Standard Cauchy distribution. The second is normalized because
For this symmetric component, let . Its density is , and hence its cumulative distribution function and quantile function are
The independent sign is positive or negative with equal probability. For one sample, use the next three independent uniforms in the given sequence and return
The Cauchy branch follows from its quantile function, while the other branch follows from the magnitude law and fair sign just derived. Independence of the branch selection and component draws therefore gives exactly the displayed mixture distribution. Discard the zero-probability endpoint draw to keep the computation finite, and use a fresh block of three uniforms for the next sample. This is the symmetric rational mixture sampler.
One can alternatively use the root solution's rejection sampling with a Standard Cauchy distribution proposal. The density ratio is , bounded by because . Accept the Cauchy draw with probability using a fresh independent uniform.