Cross-validation analysis

    Cross-validation is the action of testing each sample in the dataset, and assessing the quality of both marker selection and subsequent classification given the dataset. To perform a cross-validation, ArrayMiner iteratively removes a sample to check from the dataset, recomputes new marker genes for all classes on the basis of the reduced dataset, and then tests if the removed experiment is correctly re-classified by the resulting set of markers. This is repeated for each sample in the dataset. A red cross on the top of a sample column indicates a failed reclassification of the sample.

    To start a cross-validation, select the Master Tab and then click the Cross val button. If the button is disabled, ArrayMiner cannot perform the task (for example if only a single class is defined). After clicking the button a window will show the progress of the analysis. When the analysis is finished, a new tab appears with the result of the analysis.

Train and Test analysis

    A train and test analysis is done in two steps. ArrayMiner first identifies a discriminant set of markers for the train classes, in the same way as for a cross-validation analysis. ArrayMiner then assigns each sample in the test set of samples to a class of the train samples. A red cross on the top of a sample column indicates a failed reclassification of the sample.

    To start a Train/test analysis, select the Master Tab and then click the Train-Test button. If the button is disabled, ArrayMiner cannot perform the task (for example if there is no training set defined, or if only one class is defined). After clicking the button a window will show the progress of the analysis. When the analysis is finished, a new tab appears with the result of the analysis.

Learn more on how to specify class attributes here from the Master Tab.