The Beta-PERT methodology allows to parametrize a generalized Beta distribution based on expert opinion regarding a pessimistic estimate (minimum value), a most likely estimate (mode), and an optimistic estimate (maximum value). The betaPERT
function incorporates two methods of calculating the parameters of a Beta-PERT distribution, designated "classic"
and "vose"
.
betaPERT(a, m, b, k = 4, method = c("classic", "vose"))
## S3 method for class 'betaPERT'
print(x, conf.level = .95, ...)
## S3 method for class 'betaPERT'
plot(x, y, ...)
a
|
Pessimistic estimate (Minimum value) |
m
|
Most likely estimate (Mode) |
b
|
Optimistic estimate (Maximum value) |
k
|
Scale parameter |
method
|
"classic" or "vose" ; see details
|
x
|
Object of class betaPERT
|
y
|
Currently ignored |
conf.level
|
Confidence level used in printing quantiles of resulting Beta-PERT distribution |
...
|
Other arguments to pass to function print and plot
|
The Beta-PERT methodology was developed in the context of Program Evaluation and Review Technique (PERT). Based on a pessimistic estimate (minimum value), a most likely estimate (mode), and an optimistic estimate (maximum value), typically derived through expert elicitation, the parameters of a Beta distribution can be calculated. The Beta-PERT distribution is used in stochastic modeling and risk assessment studies to reflect uncertainty regarding specific parameters.
Different methods exist in literature for defining the parameters of a Beta distribution based on PERT. The two most common methods are included in the BetaPERT
function:
Classic
The standard formulas for mean, standard deviation, \(\alpha\) and \(\beta\), are as follows
\[mean = (a + k*m + b) / (k + 2)\]
\[sd = (b - a) / (k + 2)\]
\[\alpha = { (mean - a) / (b - a) } * { (mean - a) * (b - mean) / sd^{2} - 1 } \]
\[\beta = \alpha * (b - mean) / (mean - a)\]
The resulting distribution is a 4-parameter Beta distribution: \(Beta(\alpha, \beta, a, b)\).
Vose
Vose (2000) describes a different formula for $$:
\[(mean - a) * (2 * m - a - b) / { (m - mean) * (b - a) }\]
Mean and β are calculated using the standard formulas; as for the classical PERT, the resulting distribution is a 4-parameter Beta distribution: \(Beta(\alpha, \beta, a, b)\).
Note: If \(m = mean\), \(\alpha\) is calculated as \(1 + k/2\), in accordance with the mc2d package (see ‘Note’).
A list of class "betaPERT"
:
alpha
|
Parameter \(\alpha\) (shape1) of the Beta distribution |
beta
|
Parameter \(\beta\) (shape2) of the Beta distribution |
a
|
Pessimistic estimate (Minimum value) |
m
|
Most likely estimate (Mode) |
b
|
Optimistic estimate (Maximum value) |
method
|
Applied method |
Available generic functions for class "betaPERT"
are print
and plot
.
The mc2d package provides the probability density function, cumulative distribution function, quantile function and random number generation function for the PERT distribution, parametrized by the "vose"
method.
Classic: Malcolm DG, Roseboom JH, Clark CE, Fazar W (1959) Application of a technique for research and development program evaluation. Oper Res 7(5):646-669. http://dx.doi.org/10.1287/opre.7.5.646
Vose: David Vose. Risk analysis, a quantitative guide, 3rd edition. Wiley and Sons, 2000.
betaExpert
for modelling a standard Beta distribution based on expert opinion
## The value of a parameter of interest is believed to lie between 0 and 50
## The most likely value is believed to be 10
## .. Classical PERT
betaPERT(a = 0, m = 10, b = 50, method = "classic")
#> method alpha beta a b mean median mode var 2.5% 97.5%
#> 1 classic 1.968 4.592 0 50 15 13.93669 10 69.44444 2.247894 33.37723
## .. Plot method
plot(betaPERT(a = 0, m = 10, b = 50, method = "classic"))
## .. Vose parametrization
betaPERT(a = 0, m = 10, b = 50, method = "vose")
#> method alpha beta a b mean median mode var 2.5% 97.5%
#> 1 vose 1.8 4.2 0 50 15 13.83361 10 75 1.960636 34.15672