Skip to contents

It works for any univariate distribution

Source: vignettes/it_works_for_any_univariate_distribution.Rmd
it_works_for_any_univariate_distribution.Rmd

It works for any univariate distribution

Although most examples presented on this site and in the function documentation involve densities implemented in R, the accept_reject() function is capable of generating samples from any univariate distribution, whether it is in R or not, or even one that has not yet been invented. In other words, as long as you have a density, just write the density and pass it to the accept_reject() function.

Modified Beta Weibull distribution

The AcceptReject package is designed to be generic and timeless. Consider, for example, the family of Modified beta distributions, proposed in the paper Modified beta distributions. This is a family of probability distributions, as it is possible to generate various probability density functions through the proposed density generator, whose general density function is defined by:

\[f_X(x) = \frac{\beta^a}{B(a,b)} \times \frac{g(x)G(x)^{a - 1}(1 - G(x))^{b - 1}}{[1 - (1 - \beta)G(x)]^{a + b}},\] with \(x \geq 0\) and \(\beta, a, b > 0\), where \(g(x)\) is a probability density function, \(G(x)\) is the cumulative distribution function of \(g(x)\), and \(B(a,b)\) is the beta function.

The authors present a quantile function in closed form that depends on the quantile of the base distribution, that is, \(G^{-1}(u) = x\) is known. But here, this is not the case. We will use the accept_reject() function to generate samples from the Modified Beta - \(G\) distribution, which is a distribution that is not natively in R and could eventually be implemented by the package user.


Got the idea?
Got the idea?


The following code implements the density (derived in \(x\) from the Modified Beta - \(G\)), for any \(G\). It is important to note that the user could implement it in another way. What matters is that the user has a function that implements the probability density function.

The user could, if desired, implement it directly for a specific \(G\), for example, Weibull. Here, it will be implemented for any \(G\).

Always remember, the accept_reject() function only needs the argument f to be a function that implements the probability density function in the continuous case or a probability mass function in the discrete case.

library(numDeriv)

pdf <- function(x, G, ...){
  numDeriv::grad(
    func = function(x) G(x, ...),
    x = x
  )
}

# Modified Beta Distributions
# Link: https://link.springer.com/article/10.1007/s13571-013-0077-0
generator <- function(x, G, a, b, beta, ...){
  g <- pdf(x = x, G = G, ...)
  numerator <- beta^a * g * G(x, ...)^(a - 1) * (1 - G(x, ...))^(b - 1)
  denominator <- beta(a, b) * (1 - (1 - beta) * G(x, ...))^(a + b)
  numerator/denominator
}

# Probability density function - Modified Beta Weibull
pdf_mbw <- function(x, a, b, beta, shape, scale)
  generator(
    x = x,
    G = pweibull,
    a = a,
    b = b,
    beta = beta,
    shape = shape,
    scale = scale
  )

# Checking the value of the integral
integrate(
  f = function(x) pdf_mbw(x, 1, 1, 1, 1, 1),
  lower = 0,
  upper = Inf
)
#> 1 with absolute error < 5.7e-05

Notice that the pdf_mbw() integrates to 1, being a probability density function. Thus, the generator() function generates probability density functions from another distribution \(G_X(x)\). In the case of the code above, the Weibull cumulative distribution function was assigned to the generator() function, which could be any other. Note also that I am deriving numerically so that the user does not need to implement the probability density, which involves a somewhat larger expression.

Note: You need to understand that all the code above is a programming strategy, but if you don’t understand it very well, that’s okay, you just need to implement the probability density function that needs to generate the observations, in the way you already know how to do. Ok? 🤔



Now that we have the probability density function with the Modified Beta Weibull density function - pdf_mbw(), let’s use the accept_reject() function to generate observations of a sequence of independent and identically distributed (i.i.d.) random variables with the Modified Beta Weibull distribution.

library(AcceptReject)
#> 
#> Attaching package: 'AcceptReject'
#> The following object is masked from 'package:stats':
#> 
#>     qqplot

# True parameters
a <- 10.5
b <- 4.2
beta <- 5.9
shape <- 1.5
scale <- 1.7

set.seed(0)
x <- 
  accept_reject(
    n = 1000L,
    f = pdf_mbw,
    args_f = list(a = a, b = b, beta = beta, shape = shape, scale = scale),
    xlim = c(0, 2.5)
  )

# Plots
plot(x)


Did you notice how easy it is to use the AcceptReject package? Just implement the probability density function and pass it to the accept_reject() function. It doesn’t matter how you implemented it; what matters is that you have the function to pass to the f argument of accept_reject().


Modified Beta Gamma distribution

In a very simple way, we can generate data for any random variable \(X \sim MBG\). Here, we will generate data for a sequence of i.i.d random variables where the base distribution is the gamma distribution. So,

# Probability density function - Modified Beta Gamma
pdf_mbg <- function(x, a, b, beta, shape, scale)
  generator(
    x = x,
    G = pgamma,
    a = a,
    b = b,
    beta = beta,
    shape = shape,
    scale = scale
  )

# True parameters
a <- 1.5
b <- 3.1
beta <- 1.9
shape <- 3.5
scale <- 2.7

set.seed(0)
x <- 
  accept_reject(
    n = 1000L,
    f = pdf_mbw,
    args_f = list(a = a, b = b, beta = beta, shape = shape, scale = scale),
    xlim = c(0, 3.5)
  )

# Plots
plot(x)


Did you like it?
Did you like it?