eval.estimation
Evaluate an estimation method
Description
Evaluate an estimation method.
Usage
eval.estimation( model, replications=100, seed=NULL, quiet=F,
simulation.args=NULL,
estimation='black.box', estimation.args=list(verbose=F),
criterion ='parms', criterion.args =NULL)
Required Arguments
Optional Arguments
- replications
-
The number of simulations.
- seed
-
The starting .Random.seed.
- quiet
-
If T then no information is printed during estimation.
- simulation.args
-
A list of any arguments to pass to simulate.
- estimation
-
A character string indicating the estimation routine to use.
- estimation.args
-
A list of any arguments to pass to the estimation routine.
- criterion
-
A function to apply to the results of estimation to extract the information which is to be retained.
- criterion.args
-
A list of any arguments to be passed to the criterion function.
Value
A list with $result of length replications, each element containing the
results of criterion(estimation(simulate(model))). Other elements of
the list contain information from the supplied arguments.
Details
estimation.args and criterion.args should be NULL if no args are needed.
If model is an object of class 'estimation.evaluation' or 'simulation'
then the model and the seed!!! are extracted so the evaluation will be
based on the same generated sample.
criterion can be in { 'parms', 'roots', 'TSmodel', 'TSestModel'}
With the default (parms) and $model the other criteria can be reconstructed
when the estimation method gets the correct form for the model. ( This
is not usually the case with the default method 'black.box'.)
This is done by via the generic functions roots, TSmodel and TSestModel.
If criterion = 'roots' then criterion.args= list(verbose=F) is advised.
example simulation.args=list(sampleT=100, sd=1.5)
See Also
Examples
z <- eval.estimation(model)
return to Table of Contents