The Graphics tab displays the graphical output from morphometric analyses. The items in this tab are themselves arranged as one or more tabs, and some of these may in turn contain several different graphs arranged as multiple tabs. For example, in the screen shot below, the Graphics tab contains the output from a single principal component analysis, for which the tab "PC shape changes" is currently active.
Clicking on a tab activates the respective graph. For additional control of the output, the popup menu of the active graph can be invoked. The contents of the popup menu depend on the particular type of graph that is displayed.
For the principal component analysis shown in the screen shot above, the user can select the principal component to be displayed, the magnitude of the shape change shown, aspects of the orientation of the landmark configuration in the graph window, as well as the type of graph used to display the shape change.
A wide range of different graphs are produced by various analyses. These can be changed or customized by invoking the popup menu. The specific options listed in the menu depend on the type of graph.
The colors of the graphs have been chosen so that different types of objects have different colors. Usually, the numbers of the landmarks are indicated in red, and features of the graphs are in dark and light blue. These colors are meant to be changed according to the user's personal taste (export as SVG file and open the graph in your favorite vector graphics program). Many programs, allow the user to select all items of a particular color. For instance, to get rid of all the landmark numbers in Adobe Illustrator, select one of them, then use Select > Same > Fill Color and then press delete.
For illustrating shape changes, MorphoJ offers four main types of graphical display: lollipop graphs, transformation grids, warped outline drawings and wireframe graphs. These are illustrated for one particular shape change below. All four types have advantages and shortcomings (for a discussion, see Klingenberg 2013).
All four types of graphs illustrate the shape change from a starting shape, usually the mean shape in the sample, by a certain amount that can be changed by selecting a different scale factor (using the popup menu of the graph). The interpetation of the scale factor depends on the analysis. For instance, for principal components, this amount is the magnitude of the shape change as a Procrustes distance (because the PC vectors themselves are scaled by convention to a length of 1.0). For other types of analysis, the scaling is different. For some analyses (e.g. PCA), the magnitude of the shape change is arbitrary, and it is therefore usually given as a Procrustes distance. For other analyses (e.g. regression), the magnitude of the shape change is fixed: in this case, the scale factor is used to determine by how much the shape change is exaggerated.
A lollipop graph shows the shifts of landmark positions with straight lines. Each line starts with a dot at the location of the landmark in the starting shape (often a mean shape, etc.). The length and direction of the line indicates the movement of the respective landmark from the starting shape to the target shape (e.g. the mean shape plus the shape change that corresponds to an increase of 0.1 units of Procrustes distance in the direction of the PC1).
Lollipop graphs are available both for 2D and 3D data. For 3D data, three different views can be requested by using Choose Axes To Display from the popup menu of the graph.
Transformation grids show the shape change as a deformation of a rectangular grid using the thin-plate spline. The graph shows the same dots and lines as a lollipop graph, but adds the defomed grid with the same scale factor. The total shape change is shown (i.e. there is no decomposition into uniform and non-uniform components, etc.).
Transformation grids are only available for 2D data. The number of horizontal and vertical grid lines can be changed via the popup menu.
Warped outline drawings are similar in that they also use the thin-plate spline, but instead of a rectangular grid, an outline drawing of the biological structure is used. The starting shape is shown as a light blue outlines with hollow dots at the positions of the landmarks, and the target shape is shown as a dark blue outline with solid dots at the positions of the landmarks.
The information for the drawing must be imported into MorphoJ from a text file. For an explanation of the file format and how to import it, follow this link. Warped outline drawings are only available for 2D data.
Finally, it is possible to display the shape change as a wireframe graph. A wireframe connects selected landmarks with straight lines.
Wireframe graphs are available for 2D and 3D data. The wireframes can be defined and edited in MorphoJ by using Create or Edit Wireframe in the Preliminaries menu.
There are specific options for each of these types of graph that can be set by using Set Options for Shape Graphs in the Prelminaries menu. These options include the colors of the different elements, whether the starting shape and landmark numbers are to be displayed, etc.
A variety of analyses produce scatter plots (e.g. for principal component or canonical variate scores). The popup menu of these plots provides a variety of options for customizing the plots, for instance, coloring the points according to a classifier variable.
The identity of any specific data point can be checked by clicking on the point while holding down the Shift key. A dialog box like the following will appear:
The box contains the value of the identifier for the point on which the user clicked. If the position of the mouse click is within the diameter of multiple points, the box lists all their identifiers.
A range of other types of graphs are also produced by MorphoJ depending on the particular analyses.
The biplot is a form of graphical presentation originally developed for combining information about variables and observations in the same plot, but is used in MorphoJ only to illustrate variables that cannot be visualized as shape changes. An example is the following graph from a two-block PLS analysis in which one of the blocks consisted of six covariates:
The graph shows two PLS axes (PLS3 in the horizontal and PLS5 in the vertical direction). It can be seen that PLS3 is primarily a contrast between the variables 'Length' and 'Weight', whereas all other variables are close to zero, and may therefore be interpreted as an index of slenderness. Likewise, the PLS5 axis is mainly a contrast between 'Vertebrae' and 'Age'.
The popup menu of most graphs contains a Print entry. This will bring up the standard print dialog of the operating system, with which the graph can be directed to different printers or a PDF file, etc., depending on the particular computer.
Invoking Export Graph to File in the popup menu brings up a dialog box for saving the active graph as a graphics file.
There are the following file formats:
Use one of the vector graphics formats for editing with graphics software. This will provide the greatest flexibility for changing colors, line widths and other attributes of the graph. Most graphs are meant to be edited before publication. Therefore, different types of items in the graphs are colored differently, so that all corresponding items can be easily selected together for changing their properties or for deleting them.
The SVG vector graphics format saved by MorphoJ can be opened and
modified by the free program Inkscape, which I now use for
generating all the figures in my own research. Many web browsers
(at least the newer versions) also open SVG files for viewing. In
addition, many commercial programs such as Adobe Illustrator,
Canvas or CorelDraw can be used to open SVG files and to modify
and save graphs as vector graphics files.
Klingenberg, C. P. 2013. Visualizations in geometric morphometrics: how to read and how to make graphs showing shape changes. Hystrix 24:15–24.