The Evince Workflow

The user has entered the Evince workspace and can now begin the data exploration. This section describes a typical example how an Evince workflow can look like. The workflow requires that the user has imported the file "imports-85.txt", which can be found in the "Samples" folder in the location where Evince is installed.

1. The imported data can be seen in the data tree to the left of the workspace. Here, a DataSet is available that can be used for creating multivariate models. Start by right-clicking on the DataSet and choose create a new PCA model.

2. The model will appear below the DataSet in the data tree. Double-click on the model or click the +-sign to the left of the model in order to expand the viewing of the model. This will reveal all the statistics that are associated with the PCA model. Now, drag the score matrix, T, from the data tree onto the plot area and release it.

3. Choose to create a Scatter 2D plot from the dialog that appears. A score scatter plot will now appear in the plot area.

4. Drag the loading matrix, P, from the data tree and release it on the plot area (The plot area is covered by the score scatter plot). Choose to create a Scatter 2D plot from the dialog that appears. A loading scatter plot can now be seen in the plot area.
5. Both the score- and the loading scatter plots can now be seen in the plot area. The score scatter plot is at the bottom.
6. Right-click on the DataSet and select "New Modification" and "Category". This will create a new category for the DataSet. A category in Evince is a variable denoting to what category an observation belongs to. Each category is divided into a number of classes.
7. Choose to create classes from identifiers or data. This option will use an entire column of the imported data for the creation of the classes. In the drop list, select "fuel-type".
8. Select the score scatter plot by clicking on it. This will make the settings panel to the left available for that particular plot. Color the observations in the plot according to the category "fuel-type" by selecting [CATEGORY] in the "Color" settings.
9. Select the loading scatter plot by clicking on it. In the settings panel, select "Identifier" from the "Label" settings to label the objects in the plot.

10. Select the free hand selection from the plot toolbar. Then select a few observations in the score scatter plot.

11. Right-click on the score scatter plot and choose "Exclude" from the plot menu that appears. This will exclude the selected observations from further modeling. Now, right-click once more and choose "Apply Changes" from the plot menu. The PCA model will now be updated with the excluded observations removed from the modeling. The score scatter plot is now updated with the excluded observations gone.