Combine (version 1.0)

ncomb.prob: Combination of probability distributions.

Description

It computes the combination of probability distributions according to few parameters.

Usage

ncomb.prob(x, y = NULL, independent = !is_nothing(nsample), lower.plausible = 0, upper.plausible = 1, method.name = character(0),  nsample = numeric(0), tolerance = 0.005, arithmetic = TRUE, loss.type = c("information","quadratic","harmonic","geometric","logarithmic", "evidential"), plots = FALSE,...)

Arguments

x

List of numeric vectors of probabilities or a numeric vector of probabilities (see note).

y

A numeric vector of probabilities. Default is NULL (see note).

tolerance

Small quantity to control the failure of the method.

independent

Logical. If independent=TRUE, one weight is computed, otherwise one weight is computed per pair of elements to be combined (see note).

lower.plausible

A number within [0,1].

upper.plausible

A number within [0,1].

method.name

Names of the probabilities distributions to be combined.

nsample

Number of samples (for Monte Carlo).

arithmetic

Logical. If arithmetic=FALSE, the logarithmic approach is applied.

loss.type

specifies loss function type: "information", "quadratic", "harmonic", "geometric", "logarithmic", "evidential".

plots

option to plot (plot=TRUE) the input and output probabilities.

...

Other numeric vectors of probabilities to combine and further arguments passed to function optimize.

Value

A list:

combined

A numeric vector that results of the combination of the data.

weight

Computed weights: a numeric value (if input independent=TRUE) or vector (if input independent=FALSE) with elements in [0,1].

info

Other information.

References

Bickel, D. R. (2012). Game-theoretic probability combination with applications to resolving conflicts between statistical methods. International Journal of Approximate Reasoning, 53, 880-891.

Examples

  #Two probability distributions to combine
x1<-c(a = 0.5, b = 0.2, d = NA, e = 0, 0.3, 0.8)
x2<-c(b = 0.4, a = 0.3, e = NA, b = 0.2 , 0.13, 0.28)

z1 <- ncomb.prob(x = x1,y= x2,independent=TRUE)
z2 <- ncomb.prob(x = list(X1=x1,X2=x2),independent=FALSE)

#not yet implemented for independent=TRUE:
z3 <- ncomb.prob(x = x1,independent=FALSE,loss.type="quadratic")
z4 <- ncomb.prob(x = as.list(x1),independent=FALSE)

Run the code above in your browser using DataCamp Workspace