Generic function for the computation of convex contamination (pseudo-)distance of two probability distributions \(P\) and \(Q\). That is, the minimal size \(\varepsilon\in [0,1]\) is computed such that there exists some probability distribution \(R\) with $$Q = (1-\varepsilon)P + \varepsilon R$$
ContaminationSize(e1, e2, ...)
# S4 method for AbscontDistribution,AbscontDistribution
ContaminationSize(e1,e2)
# S4 method for DiscreteDistribution,DiscreteDistribution
ContaminationSize(e1,e2)
# S4 method for AcDcLcDistribution,AcDcLcDistribution
ContaminationSize(e1,e2)
A list containing the following components:
object of class "Distribution"
; ideal distribution
object of class "Distribution"
; 'contaminated' distribution
size of contamination
object of class "Distribution"
object of class "Distribution"
further arguments to be used in particular methods (not in package distrEx)
convex contamination (pseudo-)distance of two absolutely continuous univariate distributions.
convex contamination (pseudo-)distance of two discrete univariate distributions.
convex contamination (pseudo-)distance of two discrete univariate distributions.
Matthias Kohl Matthias.Kohl@stamats.de,
Peter Ruckdeschel peter.ruckdeschel@uni-oldenburg.de
Computes the distance from e1
to e2
respectively
\(P\) to \(Q\). This is not really a distance as it is not symmetric!
Huber, P.J. (1981) Robust Statistics. New York: Wiley.
KolmogorovDist
, TotalVarDist
,
HellingerDist
, Distribution-class
ContaminationSize(Norm(), Norm(mean=0.1))
ContaminationSize(Pois(), Pois(1.5))
Run the code above in your browser using DataLab