Visualizing the Code-Document Table: Sankey Diagram

Data Visualization is a great way to simplify the complexity of understanding relationships among data. The Sankey Diagram is one such powerful technique to visualize the association of data elements. Originally, Sankey diagrams were named after Irish Captain Matthew Henry Phineas Riall Sankey, who used this type of diagram in 1898 in a classic figure showing the energy efficiency of a steam engine. Today, Sankey diagrams are used for presenting data flows and data connections across various disciplines.

Video Tutorial: Sankey Diagrams in ATLAS.ti Windows

  • Sankey diagrams allow you to show complex processes visually, with a focus on a single aspect or resource that you want to highlight.

  • They offer the added benefit of supporting multiple viewing levels. Viewers can get a high level view, see specific details, or generate interactive views.

  • Sankey diagrams make dominant factors stand out, and they help you to see the relative magnitudes and/or areas with the largest contributions.

In ATLAS.ti, the Sankey diagram complements the Code Document Table. As soon as you create a table, a Sankey diagram visualizing the data will be shown below the table.

Sankey Diagram for the Code-Document Table

The row and column entities of the table are represented in the Sankey model as nodes and edges, showing the flow between each pair of nodes. In the Code Document table, the connecting pairs can consist of:

  • codes - documents
  • codes - document groups
  • code groups - documents
  • code groups - document groups

For each table cell containing a value, an edge is displayed between the diagram nodes. The thickness of the edges resemble the cell values of the table. Cells with value 0 are not displayed in the Sankey view.

The key to reading and interpreting Sankey Diagrams is to remember that the width is proportional to the quantity represented.

If you only want to see the table, or the Sankey diagram:

You can select what you want to see by making a selection in the ribbon at the top left-hand side. If you need more space on your screen for the diagram, you can deactivate the selection lists under the View tab.

Code-Document Table View Options

Layout

The Sankey diagram, similar to the network editor, applies a layout for its nodes and consequently to the edges connecting nodes to create an easy to comprehend visualization of the data. This basic layout places the selected entities (nodes in Sankey terminology) into vertical „layers“ with the row entities placed to the left and the column entities to the right. If nodes have incoming as well as outgoing links in the currently visible set of nodes, they will be placed in intermediate lanes or layers. In many situations, the initial layout already meets the researchers requirements. However, there are some options to modify the initial layout.

Swapping columns and rows for the table will also affect the Sankey diagram and result in a 180° rotation.

Interactivity

  • To display the quotations represented by the link between two edges, click on the connecting edge. This has the same effect as clicking a table cell. You see the quotations displayed in the Quotation Reader to the right.

  • The label of the edges show the numbers that you also see in the table cells, which are by default the number of codings in the document or document group.

  • To support finding the correct link, moving the mouse highlights the hovered edges. After selecting an edge, the unselected edges are dimmed to further increase the visibility of the edge under consideration.

  • Selecting a node will highlight this node as well as all its incoming and outgoing edges. This makes it easy to spot areas of connectivity.

  • You can also use the nodes label for selection, differing only by doing a „single-handed“ multi-selection which otherwise would require pressing the Ctrl-key.

Removing edges

To unscramble a potentially cluttered diagram, you can remove selected nodes or edges by pressing the Delete key on your keyboard. You can undo this via Ctrl-Z.

The same may be accomplished by modifying the selection lists, but this is a bit „heavier“, and it may not be clear which entities need to be deselected.

Code-Document Table Sankey Diagram Options

Fit to Window

When the Sankey diagram is created (concurrently with the respective tools’ table) it may display a bit too tiny in its reserved window pane. There are several options to fix this: Switch off the table view and the selection lists, maximize the project window or float the tool with the diagram. The Sankey diagram itself will not resize to the freed space automatically. Click button Fit-to Window to fit the current diagram to the available window area.

Refresh Layout

This will also move nodes to their original places if moved in the meantime. The edges will also be recalculated for best entanglement. This option can also be activated by double clicking the diagrams background.

Auto Layout

This option is activated by default. It will instantly recompute the diagram’s layout if:

  • Nodes are moved

  • Nodes are removed

  • Edges are removed

Respect Node Placement

You might not always be happy when the nodes you shifted to other locations are instantly moved back to their original locations. To prevent this, you either need to disable the Auto Layout or select Respect Node Placement.

Note, that the Respect Node Placement option will not let you place nodes to arbitrary positions but will find a nearest lane and relayout the diagram to make it nice looking again.

Shortcut: Double click the diagram background while pressing the left Shift key.

Rotate Diagram

Sometimes the horizontal left-to-right layout may not always give the most optimal visualization of your data (depending on available screen space, monitor size). Click this button to toggle between a vertical (top-down) and the original horizontal layout.

Shortcut: Double click the diagram background while pressing the left Ctrl key.

Display Icons

This display the entities type icons preceding the labels. This makes most sense when the diagram contains multiple types of entities as in the Code Document Table. By default, this option is not activated.

The coloring of the links and edges between the nodes can be set to:

  • Use the color of the source node (either the user provided color for codes, or the type dependent color of other entities.

  • Use the color of the target node

  • Use a color gradient starting with the source and ending with the target color. This is the default. It makes tracking the endpoint of an edge quite comfortable. Shortcut: Double click the diagram background while pressing the left Alt key.

  • Use a gray shade independent of any node coloring.

Shortcut: Double click the diagram background while pressing the left Alt key.

Dark / Light Background

This toggles a light and a dark background, just for aesthetics.

Shortcut: Double click the diagram background while pressing the left Ctrl and Shift key.

Creates an output of the current diagram respecting both the selected background, and the visual selection state of both nodes and edges. The printout will not contain any quotations. Printing can also create high quality PDFs if an appropriate printer driver is installed.

Miscellaneous

The diagram is integrated into the table tools and responds to certain options:

  • Compress, that is removing empty rows or columns, will remove corresponding elements in the diagram.

  • Clustering can reduce the number of co-occurrences and thus influences the size of the edges in the diagram.

  • Swapping rows and columns reverts the original layout by 180°.