UnitsSameCluster

UnitsSameCluster enables to control the labels of the clusters which contain predefined units

expand all in page

Syntax

  • IDXwithConsistentLabels=UnitsSameCluster(IDX,UnitsSameGroup)example
  • [IDXwithConsistentLabels, OldAndNewIndexes]=UnitsSameCluster(___)example

Description

example

IDXwithConsistentLabels =UnitsSameCluster(IDX, UnitsSameGroup) Start with labelling produced by tclustIC and produce consistent labels.

example

[IDXwithConsistentLabels, OldAndNewIndexes] =UnitsSameCluster(___) Example with detailed description of output element OldAndNewIndexes.

Examples

expand all

  • Start with labelling produced by tclustIC and produce consistent labels.

    •     Y=load('geyser2.txt');
    % A small number of subsamples just to show whow the procedure works.
    nsamp=10;
    out=tclustIC(Y,'cleanpool',false,'plots',1,'nsamp',10,'whichIC','CLACLA')
    % Make sure that units [23 54] are whenever possible respectively in
    % cluster 1 and 2
    UnitsSameGroup=[23 54];
    IDXCLAnew=UnitsSameCluster(out.IDXCLA,UnitsSameGroup);

  • Example with detailed description of output element OldAndNewIndexes.

    • Random seed to be example ro replicate the results. 

        rng(1000)
    Y=load('geyser2.txt');
    k=3;
    [out]=tclust(Y,k,0.10,10);
    % Make sure that group which contains
    % unit 10 is always labelled with number 1. Similarly,
    % make sure that the group which contains unit 12 is always labelled
    % with number 2, 
    UnitsSameGroup=[10;12];
    [idxnew, OldNewIndexes]=UnitsSameCluster({out.idx}, UnitsSameGroup);
    % In this case OldNewIndexes is equal to 
    % 3 1 
    % 3 2 
    % It means that in the first iteration labels 1 and 3 have swapped
    % while in the second iteration label 3 and 2 have swapped
    subplot(1,2,1)
    gscatter(Y(:,1),Y(:,2),out.idx)
    text(Y(UnitsSameGroup,1),Y(UnitsSameGroup,2),num2str(UnitsSameGroup))
    subplot(1,2,2)
    gscatter(Y(:,1),Y(:,2),idxnew{:})
    text(Y(UnitsSameGroup,1),Y(UnitsSameGroup,2),num2str(UnitsSameGroup))
    % Now (as is evident from the right panel) unit which contains group 10
    % has label '1' while group which contains unit 12 has label '2'.

    ClaLik with untrimmed units selected using crisp criterion
Total estimated time to complete tclust:  1.61 seconds

    Input Arguments

    expand all

    IDX — Assignment of units to groups for different values of c (restriction factor) and k (number of groups). Cell.

    Cell of size length(kk)-times length(cc), where kk is the vector which contains the number of groups which have been considered and cc is the vector which contains the values of the restriction factor. Each element of the cell is a vector of length n containing the assignment number of each unit using a particular classification model.

    Data Types: cell

    UnitsSameGroup — list of the units which must (whenever possible) have the same label. Numeric vector.

    For example if UnitsSameGroup=[20 26], means that group which contains unit 20 is always labelled with number 1. Similarly, the group which contains unit 26 is always labelled with number 2, (unless it is found that unit 26 already belongs to group 1). In general, group which contains unit UnitsSameGroup(r) where r=2, ...length(kk)-1 is labelled with number r (unless it is found that unit UnitsSameGroup(r) has already been assigned to groups 1, 2, ..., r-1).

    Data Types: integer vector

    Output Arguments

    expand all

    IDXwithConsistentLabels —cell with the same size as input cell IDX and with the same meaning of input cell IDX but with consistent labels. Cell

    Group which contains unit UnitsSameGroup(1) is labelled with number 1. In general. Group which contains UnitsSameGroup(r) where r=2, ...length(kk)-1 is labelled with number r (unless it is found that unit UnitsSameGroup(r) has already been assigned to groups 1, 2, ..., r-1).

    OldAndNewIndexes —indexes of the permutations associated with IDX{1,1}. r-by-2 matrix

    Matrix of size r-by-2 which keeps track of all the permutations which have been done. For example if OldAndNewIndexes is equal to [3, 1; 3, 2], it means that in the first iteration labels 1 and 3 have swapped, while in the second iteration label 3 and 2 have swapped. If no swapping was necessary OldAndNewIndexes is empty.

    References

    A. Cerioli, L.A. Garcia-Escudero, A. Mayo-Iscar and M. Riani (2017), Finding the Number of Groups in Model-Based Clustering via Constrained Likelihoods, Journal of Computational and Graphical Statistics, https://doi.org/10.1080/10618600.2017.1390469

    See Also

    |

    This page has been automatically generated by our routine publishFS