GammaDist {stats}R Documentation

The Gamma Distribution

Description

Density, distribution function, quantile function and random generation for the Gamma distribution with parameters shape and scale.

Usage

dgamma(x, shape, rate = 1, scale = 1/rate, log = FALSE)
pgamma(q, shape, rate = 1, scale = 1/rate, lower.tail = TRUE,
       log.p = FALSE)
qgamma(p, shape, rate = 1, scale = 1/rate, lower.tail = TRUE,
       log.p = FALSE)
rgamma(n, shape, rate = 1, scale = 1/rate)

Arguments

x, q

vector of quantiles.

p

vector of probabilities.

n

number of observations. If length(n) > 1, the length is taken to be the number required.

rate

an alternative way to specify the scale.

shape, scale

shape and scale parameters. Must be positive, scale strictly.

log, log.p

logical; if TRUE, probabilities/densities p are returned as log(p).

lower.tail

logical; if TRUE (default), probabilities are P[X \le x], otherwise, P[X > x].

Details

If scale is omitted, it assumes the default value of 1.

The Gamma distribution with parameters shape =\alpha and scale =\sigma has density

f(x)= \frac{1}{{\sigma}^{\alpha}\Gamma(\alpha)} {x}^{\alpha-1} e^{-x/\sigma}%

for x \ge 0, \alpha > 0 and \sigma > 0. (Here \Gamma(\alpha) is the function implemented by R's gamma() and defined in its help. Note that a = 0 corresponds to the trivial distribution with all mass at point 0.)

The mean and variance are E(X) = \alpha\sigma and Var(X) = \alpha\sigma^2.

The cumulative hazard H(t) = - \log(1 - F(t)) is

-pgamma(t, ..., lower = FALSE, log = TRUE)

Note that for smallish values of shape (and moderate scale) a large parts of the mass of the Gamma distribution is on values of x so near zero that they will be represented as zero in computer arithmetic. So rgamma may well return values which will be represented as zero. (This will also happen for very large values of scale since the actual generation is done for scale = 1.)

Value

dgamma gives the density, pgamma gives the distribution function, qgamma gives the quantile function, and rgamma generates random deviates.

Invalid arguments will result in return value NaN, with a warning.

The length of the result is determined by n for rgamma, and is the maximum of the lengths of the numerical arguments for the other functions.

The numerical arguments other than n are recycled to the length of the result. Only the first elements of the logical arguments are used.

Note

The S (Becker et al., 1988) parametrization was via shape and rate: S had no scale parameter. It is an error to supply both scale and rate.

pgamma is closely related to the incomplete gamma function. As defined by Abramowitz and Stegun 6.5.1 (and by ‘Numerical Recipes’) this is

P(a,x) = \frac{1}{\Gamma(a)} \int_0^x t^{a-1} e^{-t} dt

P(a, x) is pgamma(x, a). Other authors (for example Karl Pearson in his 1922 tables) omit the normalizing factor, defining the incomplete gamma function \gamma(a,x) as \gamma(a,x) = \int_0^x t^{a-1} e^{-t} dt, i.e., pgamma(x, a) * gamma(a). Yet other use the ‘upper’ incomplete gamma function,

\Gamma(a,x) = \int_x^\infty t^{a-1} e^{-t} dt,

which can be computed by pgamma(x, a, lower = FALSE) * gamma(a).

Note however that pgamma(x, a, ..) currently requires a > 0, whereas the incomplete gamma function is also defined for negative a. In that case, you can use gamma_inc(a,x) (for \Gamma(a,x)) from package gsl.

See also https://en.wikipedia.org/wiki/Incomplete_gamma_function, or https://dlmf.nist.gov/8.2#i.

Source

dgamma is computed via the Poisson density, using code contributed by Catherine Loader (see dbinom).

pgamma uses an unpublished (and not otherwise documented) algorithm ‘mainly by Morten Welinder’.

qgamma is based on a C translation of

Best, D. J. and D. E. Roberts (1975). Algorithm AS91. Percentage points of the chi-squared distribution. Applied Statistics, 24, 385–388.

plus a final Newton step to improve the approximation.

rgamma for shape >= 1 uses

Ahrens, J. H. and Dieter, U. (1982). Generating gamma variates by a modified rejection technique. Communications of the ACM, 25, 47–54,

and for 0 < shape < 1 uses

Ahrens, J. H. and Dieter, U. (1974). Computer methods for sampling from gamma, beta, Poisson and binomial distributions. Computing, 12, 223–246.

References

Becker, R. A., Chambers, J. M. and Wilks, A. R. (1988). The New S Language. Wadsworth & Brooks/Cole.

Shea, B. L. (1988). Algorithm AS 239: Chi-squared and incomplete Gamma integral, Applied Statistics (JRSS C), 37, 466–473. doi:10.2307/2347328.

Abramowitz, M. and Stegun, I. A. (1972) Handbook of Mathematical Functions. New York: Dover. Chapter 6: Gamma and Related Functions.

NIST Digital Library of Mathematical Functions. https://dlmf.nist.gov/, section 8.2.

See Also

gamma for the gamma function.

Distributions for other standard distributions, including dbeta for the Beta distribution and dchisq for the chi-squared distribution which is a special case of the Gamma distribution.

Examples

-log(dgamma(1:4, shape = 1))
p <- (1:9)/10
pgamma(qgamma(p, shape = 2), shape = 2)
1 - 1/exp(qgamma(p, shape = 1))

# even for shape = 0.001 about half the mass is on numbers
# that cannot be represented accurately (and most of those as zero)
pgamma(.Machine$double.xmin, 0.001)
pgamma(5e-324, 0.001)  # on most machines 5e-324 is the smallest
                       # representable non-zero number
table(rgamma(1e4, 0.001) == 0)/1e4

[Package stats version 4.5.0 Index]