This feature provides you with the capability to remove samples based on variables. Samples that have the same variable value (observation) as the one selected are removed. For example: users can create a new variable with the variable values of "keep" or "remove". If during sample removal, they select the new variable and the "remove" value, any sample with that variable value will be removed in the resulting experiment (leaving only the samples with the "keep" variable value).
This feature can be used to remove bad samples or restrict further analysis to a subset of their data. This feature can also be used as a complement to classification. One particular class (variable value) may cause difficulties in classifying other samples. It is occasionally beneficial to classify multiple smaller problems if attempting to classify all of the samples at once is not satisfactory. This feature can also be used in conjunction with Random Variables to create train and test data sets.
1. Click a regular gene expression dataset in the Experiments navigator. The item is highlighted.
2. Select Sample Removal from the Statistics menu. The Sample Removal dialog is displayed.
3. Select a Variable from the drop-down list.
4. Select the Value of the specified Variable that will be used to remove samples.
5. Click OK.
Samples in the dataset whose variable value for samples that correspond to the specified Variable Value are removed.
Note: you can import new variables against Sample Removal experiments. Variables are propagated upwards and downwards in the experiment tree. Descendent samples are marked as unknown if their observations for a given variable aren't unanimous.