About the Robust Toolbox interface
The graphical interface for robust calibration have been developed as a result of cooperation between the Department of Chemometrics at the Silesian Univeristy (Poland) and Micro and Trace Analysis Centre, ChemometriX at the Univeristy of Antwerp (Belgium).
- Dr. M. Daszykowski
- Dr. S. Serneels
- Prof. B. Walczak
- Prof. P. Van Espen
- Prof. C. Croux
- M. Daszykowski is thankful to the Foundation for Polish Science for the financial support.
- Sven Serneels: Research financed by a PhD grant of the Institute for the Promotion of Innovation through Science and Technology in Flanders (IWT-Vlaanderen).
Graphical interface for Robust Calibration
The graphical interface is equipped with a collection of m-files and its routines are freely available. These
were developed under MATLAB 6.5, one of the most popular programming environments among chemometricians.
The graphical interface allows a user to apply the implemented methods in an easy way and it gives reach possibilities
to visualize the obtained results. Several useful features such as interactive numbering of
the displayed objects on a plot, viewing the content of the data, easy transfer the data from the Toolbox to
the file and between the Toolbox, and MATLAB workspace are also implemented.
M. Daszykowski, S. Serneels, K. Kaczmarek, P. Van Espen, C. Croux, B. Walczak, TOMCAT: a MATLAB toolbox for multivariate calibration techniques, Chemometrics and Intelligent Laboratory Systems, 85 (2007) 269-277.
Software specification and requirements
The routines for robust calibration as well as the graphical interface has been developed under MATLAB 6.5 (release 13). Dependent on user's knowledge about MATLAB, either the m-files or the graphical interface can be used. The interface routine, called 'TOMCAT', requires for its proper functioning a specific structure of the catalogues where the routines are located. Apart from the main interface, several smaller interfaces are also designed enabling to define inputs and custom options for the applied method. The following catalogues with m-files are automatically made when the zip file is extracted: 'Calibration', 'Data', 'Interfaces', 'PCA', 'Preprocessing' and 'Subset Selection'.
Collection of the implemented methods
The methods which are in the graphical interface can be divided into:
- column centering
- Standard Normal Variate (SNV)
- centering by median
- centering by robust median (L1-median)
- median of absolute deviation
- robust standardization by Sn and Qn estimators
- Projection methods
- Principal Component Analysis (PCA)
- Robust Principal Component Analysis (RPCA)
- Classical linear calibration methods
- Continuum Power Regression (CR)
- Partial Least Squares (PLS): WIM and SIM variants
- Robust linear calibration methods
- Robust Continuum Regression (RCR)
- Partial Robust M-Regression (PRM)
For a more detailed overview of methods inputs and outputs pleae visit 'Overview of m-files' section.
Starting the Toolbox
To initialize the graphical interface, the current directory in MATLAB should be changed to the directory where the graphical interface is located, by using a command 'cd' (cd D:/Matlab/RobustToolbox). Typing the command 'TOMCAT' in the MATLAB workspace executes the graphical interface. When the graphical interface starts, the information about location of the required m-files is added into the MATLAB paths automatically. Therefore, the user does not have to do it before using the graphical interface.
Loading the data
To load the data into the graphical interface, select from the upper window menu a folder 'Import' and chose a location of the data: 'Import data from *.mat file' or 'Import data from workspace'. The last option is available only if the data are present in the MATLAB workspace. When the file or the workspace contains several variables then a user is asked to select necessary variables that should be imported to the graphical interface. What should be stressed is that the graphical interface can only handle MATLAB files, i.e., files with mat extension, and variables such are vectors and matrices, being double arrays. When the variables are loaded into the graphical interface the information about them including their names and sizes (number of objects and variables) is displayed in 'Data in the Toolbox' panel. At any time a user can easily export the obtained results either to a mat file that can be later used or to the MATLAB workspace, by selecting from the upper window menu an option 'Export' and choosing the destination: 'Export data to *.mat file' or 'Export data to work space'. This gives a possibility to apply other methods on the saved or exported data that are not available in the graphical interface or to create custom plots. Additionally, we have included in the graphical interface an option allowing browsing the content of the data. By double click on the left mouse button the content of the highlighted variable in the 'Data in the Toolbox' panel its content is displayed as a data sheet in MATLAB Array Editor (see Fig. 2). Simultaneously, the selected variable is exported to the workspace. However, due to the limitation of the Array Editor, only the variables containing less than 65536 elements can be viewed.
Deleting the data
The variables in the graphical interface and/or the workspace can be erased by selecting from the upper window menu a folder 'Clear' and then 'Clear data from Toolbox' or 'Clear data from workspace'. Erasing the variables from the graphical interface will result in disabling previously active options, and at this stage a user is allowed to import the data only.
Selecting the data
Once the data are loaded into the graphical interface dependent on a purpose it is necessary to select a set of independent variables, X, and optionally independent variables, y (both variables should have the same number of objects, and X should be a matrix, otherwise an error dialog box appears on the screen). This can in done using the buttons from 'Select model set and test set' panel. There are four buttons for selecting model set data independent variables (denoted as Xm), dependent model set variable (denoted as Ym) and if necessary, test set independent variables (denoted as Xt) and dependent test set variable (denoted as Yt). As soon as the model set independent variables are selected a new variable, Xm, appears in the 'Data in the Toolbox' panel and the 'Principal Component Analysis' panel becomes active allowing to apply PCA and robust PCA on this data. Anytime, the user may remove a selected variable, by clicking on the corresponding button 'Remove' and to re-select a new one. In the same way the dependent model set variable can be chosen, and then, four additional panels become active ('Subset selection', 'Classical linear calibration', 'Robust linear calibration' and 'Non-linear PLS').
The Toolbox as well as the M-files code belong to the holder of the copyrights. It is made public and free of charge for academic
use in the hope that it will be useful, but WITHOUT ANY WARRANTY. The users of our Toolbox are kindly ask to cite the following publication:
Michał Daszykowski, Sven Serneels, Krzysztof Kaczmarek, Piet Van Espen, Christophe Croux, Beata Walczak, TOMCAT: a MATLAB toolbox for multivariate calibration techniques, Chemometrics and Intelligent Laboratory Systems, 85 (2007) 269-277.
In case of doubt, please contact the holders of the copyrights either:
Department of Chemometrics
The University of Silesia
Institute of Chemistry
9 Szkolna Street
- Micro and Trace Analysis Centre