TLE: Trimmed Likelihood Estimator

Description

TLE implements a general framework for robust fitting of finite mixture models. Parameter estimation is performed using the EM algorithm.

Currently two model drivers are inluded: flexmix.Density (flexmix.Enstimate) for gaussian, poisson and binomial regression models and FLXmclust.Density (FLXmclust.Estimate) for model based clustering.

Usage

TLE(formula,family,data,kStar=NULL, kTrim=NULL, nit = 10, msglvl = 0,  result = NULL, cit = 9, test = NULL,nc=1, Density, Estimate, ...)

Arguments

formula

An object of class formula.

family

The family to be used.

data

Data frame containing the x and y variables with an optional attribute family being either gaussian,poisson or binomial

kStar

k*- size of the initial random subsample

kTrim

Trimming parameter: size of the C-steps random subsample

nit

Number of iterations

msglvl

Level of messages

result

Restart/continuation information

cit

Number of iterations in refinement step

test

Expected true loglikelihood of the model; procedure will be stopped if reached.

Number of components.

Density

Density function of type - function(data,solution,model,family,...)

Estimate

Specific estimation procedure interface: function(data,ind,model,family,...)

...

Arguments to be passed to methods Estimate and Density

Value

Returns an object of class TLE.

References

N. Neykov, P. Filzmoser, R. Dimova, and P. Neytchev. Robust fitting of mixtures using the trimmed likelihood estimator. Computational Statistics and Data Analysis, Vol. 17(3), pp. 299-308, 2007.

Examples

Run this code

data(gaussData)
# This example takes some seconds:
#est.tle	<- TLE(y~x,"gaussian",data=gaussData,nit=4, msglvl=1, cit=3, 
#           Density=flexmix.Density, Estimate=flexmix.Estimate, nc=2)

# Plot the 2-dimensional data			
#tleplot(est.tle, gaussData)