MATLAB: Do the observation labels not appear on the graph created using the BIPLOT function in the Statistics Toolbox

analysiscomponentspcaplotprincipalStatistics and Machine Learning Toolboxtext;

When I use the BIPLOT function with the commands:

oslabels = {'1','2','3'}
biplot(coefs, 'Scores', scores, 'ObsLabels', oslabels)

the resulting figure is missing the labels contained in "oslabels."

Best Answer

This is an error within the documentation for the Statistics Toolbox in the usage of the BIPLOT function. The documentation should read as follows:

 BIPLOT(COEFS, ..., 'Scores', SCORES, 'ObsLabels', OBSLABS) uses the text in the character array or cell array OBSLABS as observation names when displaying data cursors.

This change has been incorporated into the documentation of the Statistics Toolbox 5.1 (R14SP3). For previous releases, read below for any additional information:

The labels are not supposed to display on the plot, because that would usually make the plot very cluttered. However the labels will show up when you use data cursors.

Attached is an example that shows how to have all the labels show up.

Related Solutions

MATLAB: Where can I find more information about Breiman’s “random forest” method used by TREEBAGGER in Statistics Toolbox 7.1 (R2009a)

This enhancement has been incorporated in Release 2010a (R2010a). For previous product releases, read below for any possible workarounds:

Breiman's "random forest" method was introduced in Statistics Toolbox 7.1 (R2009a). You get random forest by default from TreeBagger. You do not need to do anything.

"Random forest" means that variables for decision splits are chosen at random for each decision tree. This is what "NVarToSample" option is for and by default it activates random selection of variables. If you want to turn off random forest, you need to set "NVarToSample" to "all".

Additional information about random forest can be found from an internet searching on "breiman random forests"

MATLAB: Obslabel error when using biplot

Hi,

There seems to be a mismatch between number of Observation Labels (chemical names) and number of rows in ‘score’ matrix. Number of Observation Labels must be same as number of rows in ‘score’ matrix. In question it is mentioned that number of variables is 14, but ‘Pah’ has 15 labels. To avoid the mismatch please check the size of ‘score’ matrix and provide equal number of labels in ‘Pah’ variable.

Also, I created a sample ‘spw’ variable of size 15x9 with random numbers and tested in the code as provided by you. It seems to work fine.

spw = randn(15,9);%Random numbers
Z = zscore(spw); % Standardized data
[coefs,score,v,g,ex] = pca(Z);
vbls = {'1','2','3','4','5','6','7','8', '9'}; % Array of variable labels for houses
Pah={'Naph','2Methyl','1Methyl','Acy','Ace','Flu','Phe','Ant','Flt','Pyr','BaAnt', 'Chr', 'BbFlu','BkFLu','BePyr'};% chem names
h6 = biplot(coefs(:,1:2),'Scores',score(:,1:2),'ObsLabels',Pah, 'VarLabels', vbls);

Hope this helps!

Best Answer

Related Solutions

MATLAB: Where can I find more information about Breiman’s “random forest” method used by TREEBAGGER in Statistics Toolbox 7.1 (R2009a)

MATLAB: Obslabel error when using biplot

Related Question