tclustfsda.object {fsdaR}

Objects returned by the function tclustfsda

Description

An object of class tclustfsda.object holds information about the result of a call to tclustfsda.

Value

The functions print() and summary() are used to obtain and print a summary of the results. An object of class tclustfsda is a list containing at least the following components:

call

the matched call

muopt

a k-by-p matrix containing cluster centroid locations. Robust estimate of final centroids of the groups

sigmaopt

a p-by-p-by-k array rray containing estimated constrained covariance for the k groups

idx

a vector of length n containing assignment of each unit to each of the k groups. Cluster names are integer numbers from 1 to k. 0 indicates trimmed observations.

size

a matrix of size (k+1)-by-3. The 1st col is sequence from 0 to k (cluster name); the 2nd col is the number of observations in each cluster; the 3rd col is the percentage of observations in each cluster.

Remark: 0 denotes unassigned units.

postprob

n-by-k matrix containing posterior probabilities. postprob[i, j] contains posterior probabilitiy of unit i from component (cluster) j. For the trimmed units posterior probabilities are 0.

emp

"Empirical" statistics computed on final classification. When convergence is reached, emp=0. When convergence is not obtained, this field is a list which contains the statistics of interest: idxemp (ordered from 0 to k*, k* being the number of groups with at least one observation and 0 representing the possible group of outliers), muemp, sigmaemp and sizemp, which are the empirical counterparts of idx, muopt, sigmaopt and size.

MIXMIX

BIC which uses parameters estimated using the mixture loglikelihood and the maximized mixture likelihood as goodness of fit measure.

Remark: this output is present just if mixt > 0.

MIXCLA

BIC which uses parameters estimated using the mixture loglikelihood and the maximized mixture likelihood as goodness of fit measure.

Remark: this output is present just if mixt > 0.

CLACLA

BIC which uses the classification likelihood based on parameters estimated using the classification likelihood.

Remark: this output is present just if mixt > 0.

notconver

number of subsets without convergence

bs

a vector of length k containing the units forming initial subset associated with muopt.

obj

value of the objective function which is minimized (value of the best returned solution).

equalweights

if equalweights=TRUE means that in the clustering procedure we (ideally) assumed equal cluster weights else (equalweitghts=FALSE means that we allowed for different cluster sizes.

h

number of observations that have determined the centroids (number of untrimmed units).

fullsol

a vector of size nsamp which contains the value of the objective function at the end of the iterative process for each extracted subsample.

X

the original data matrix X.

Examples

 ## Not run: 
 data(hbk, package="robustbase")
 (out <- tclustfsda(hbk[, 1:3], k=2))
 class(out)
 summary(out)
 
## End(Not run)