The aim of
package "
distrMod" is
to introduce parametric models built up on package
"distr" : The key class
we introduce is
"ParamFamily"
bringing
together parameters and distributions; most importantly, we
have a mapping function
modifyParam
mapping a parameter theta to the
corresponding distribution member of the parametric family. Based on
this we introduce
MCEs
(Minimum Criterion Estimators) which
are defined as minimizing a criterion function at the data within the
parametric family.
These are realized as a function taking as arguments
data,
ParamFamily,
criterion, the latter being the
criterion function with arguments
data and
theta measuring the criterion at
parameter value theta.
Particular cases are
- Maximum Likelihood Estimators (with the negative (log)likelihood
as criterion),
- Minimum distance estimators (distance of the empirical
distribution to the members of the parametric family as criterion);
we have realized
- Minimum Kolmogoroff distance Estimators
- Minimum Cramér von Mises distance Estimators
- Minimum Hellinger distance Estimators
- Minimum Total Variation distance Estimators
the criterion function is then optimized using the R-functions
optim,
optimize, respectively.
MCEs can (but do not need to) use method dispatch; this way particular
methods beyond numerical optimization can be used
without changing "
distrMod"
code