Contents - Index
(This feature is only available in GenEx Enterprise)
Potential curves is a predictive application of principal component analysis. To use potential curves, the data must have a training set divided into groups and there must also be test samples.These settings are made in the Data manager. Potential curves performs PCA but only on the training data. The principal components are then used to classify the test data. Furthermore, iso-probability potential curves are calculated for each data group to aid classification.
Open the PCA tab among the analysis tabs in the top of the main window, and press the P-Curve button to load the analysis into the Control panel.
Adjust the number of levels in the No. of levels drop-down list, which defines how many lines will be used when drawing the potential curve. You can also adjust the Fitness of the curves. The potential curves are drawn in the colors that you define in the Control panel or in the Data Manager under the Color & Symbols tab for Values.
The result is a PCA plot showing the samples and the potential curves for each defined group. In the example below two groups were defined (positive in red, negative in green), and 9 samples were classified as test samples in the Data manager. The test samples are labeled in the plot.
For the test data set, probabilities are calculated (given in percent) for which group they belong to. The coordinates for the principal components as well as the probabilities are shown in a table.
M. Forina, C. Armanino, R. Leardi and G. Drava (1991). A class-modelling technique based on potential functions. Journal of Chemometrics, vol 5, 5, pp 435-453.
X. Tomas and J. M. Andrade (1999). Application of simplified potential curves to classification problems.Quimica Analitica, 18, pp 225-231.