This page summarizes changes in the versions of MorphoJ. This listing includes the changes that might affect results from analyses, but many minor bug fixes that have no major effects on the functionality of the program are not listed here.
This release contains an addition to the interaction of the MorphoJ software with the operating system (Windows and Mac OS): upon installation of the software, the .morphoj file extension is associated with MorphoJ. Double-clicking a file with a MorphoJ project previously saved by the software, which has the .morphoj file extension, will start MorphoJ and load the project contained in the file. If MorphoJ is already open, the response to double-clicking in a file depends on the operating system. In MS Windows, this will open a new copy of MorphoJ with the respective project. In Mac OS, a dialog box will appear and ask the user whether to replace the project open in MorphoJ by the one in the file that was double-clicked. For Ubuntu/Linux, unfortunately, the file association appears not to work at all.
For scatter plots, a new option available from this version is to draw convex hulls. Convex hulls are implemented in a similar way to confidence ellipses, including grouping by a classifier variable, and can be requested via the popup menu of applicable scatter graphs.
This release also contains several fixes of minor bugs, including
one concerning the import of data from Morphologika files (message
on Morphmet2 by Helmi Hadi from 20 May 2024).
This release is primarily a change in the packaging and delivery of the MorphoJ software, with some minor additions to functionality and bug fixes. From this release on, MorphoJ comes as a self-contained package including its own OpenJDK runtime environment. It therefore runs without an external installation of other Java software. For this reason, the download and installation is also somewhat different from previous releases for all operating systems. This version of MorphoJ can be installed on the same machine as any of the older versions (1.07a and older).
This release includes changes in the statistical tests associated with canonical variate analysis (CVA). As a new addition, there is a global permutation test of group effects against the null hypothesis of no differences among any group means at all. In addition, there is still the option of running pairwise permutation tests between group averages. All these tests use Pillai's trace and Goodall's F as the test statistics.
This update also contains long-overdue fixes to various bugs. Examples are the updating of output covariance matrices from Procrustes ANOVAs, and a bug for differences between subgroups, where an error message was erroneous and displayed the wrong group name(s) as being associated with incomplete data. Thanks to users for their patience.
From this update on, TPS files can contain records with no landmarks (i.e. the LM= keyword is set to LM=0). For instance, these correspond to images that could not be digitized. MorphoJ tolerates these records: it lists them in the Reports tab during the import process, but otherwise ignores them. All other observations in the same TPS file must have the same number of landmarks.
This update is a switch to Oracle Java, version 8. This is not a substantial change for Windows and Unix/Linux systems, but it is a significant change for Mac OS X, where the legacy version of Apple Java 6 no longer works under recent upgrades of the Mac OX operating system, and therefore has been abandoned in favour of Oracle Java 8, which seems to work well. Therefore, Mac OS X users now should download the current version of Java from www.java.com just as Windows and Unix/Linux users.
The changes in this version mean that the structure of files inside the MorphoJ folder in the installation directory is incompatible with previous versions. If you want to install this version in the same place as a previous version (probably a good idea), you first need to delete the previous installation folder. If you don't do this, you will get an error message during the installation process.
In accordance with these changes, there are some changes in the deployment and distribution process too, as well as updates in the software used for packaging and distributing MorphoJ. Most of these changes should not be apparent to users.
One change that will be apparent is the export of graphical output. The new version abandons the use of the FreeHEP library for vector graphics, which has caused some difficulties. The recommended way to export graphics is as .SVG files, a standard file format for vector graphics that can be opened and further edited by many free and commercial graphics software packages.
The update also introduces a new utility function serves to compute differences between subgroup averages within multiple groups (Difference of Averages by Groups in the Preliminaries menu). For instance, in a study of the evolution of sex dimorphism in multiple species, this function can produce a new dataset in which the observations are the differences between the male and female averages for each species.
For reading data from TPS files, MorphoJ recognizes missing landmarks that are indicated by coordinate values of -1.0 in the TPS file (since the update of tpsDig2 to version 2.18, see MorphMet posting by Jim Rohlf on 1 April 2015). By default, the observations with missing landmarks are excluded from the dataset, but the observations can be restored by excluding the affected landmarks (see Select Landmarks in the Preliminaries menu).
The new version also includes fixes for a number of bugs, which were reported to me by various users -- they were that many over the years since the last release that I've lost track of them...
This release contains a bug fix concerning the reading of Nexus files (Import Phylogeny File). If the file contains numbers for internal nodes, the previous versions tried to translate those numbers and insert taxon names as node labels, which could lead to crashes when mapping morphometric data onto the phylogenies. The new version handles the labels for internal nodes of the phylogeny correctly (i.e. it does not change them). Thanks to Ashley Hammond for alerting me to the problem.
The new version also contains fixes for various minor but annoying bugs, including one that caused an error when mapping a single covariate onto a phylogeny. Thanks to several participants of the workshop in Els Hostalets for showing me the bugs.
Another change concerns the test for hypotheses of modularity (Modularity: Evaluate Hypothesis). Previous versions reported the number and proportion of landmark partitions with RV coefficients smaller than the one for the landmark partition specified by the hypothesis of modularity. The new version reports the number and proportion of partitions with RV coefficients smaller than or equal to that of the hypothesis. This is more in keeping with conventional statistical tests.
Finally, a further minor bug fix concerns the importing of covariance matrices and converting them back to the coordinate system of landmarks. If the data have object symmetry, a bug caused the symmetry properties to be incorrectly transferred. Thanks to Elis Damasceno for telling me about the problem.
This release contains an addition to the procedure for principal component analysis. The output now contains some integration statistics in addition to the standard PCA outputs (variance of eigenvalues, variance of eigenvalues scaled by the total variance, and scaled to an interval between 0 and 1). Note that these statistics are only displated for PCAs run under the new version of MorphoJ.
There are also a few bug fixes. The most noticeable concered importing data from TPS files. Importing data from TPS files containing records starting with LM=0 (no landmarks for an object) produced an erroneous record in the dataset. The new version of MorphoJ is issuing an error message and aborts the data import in this situation. Thanks to the participants of the Plant Morphometrics workshop in Els Hostalets de Pierola for alerting me to this bug.
This release contains a bug fix in the test for hypotheses of
modularity (Modularity:
Evaluate Hypothesis). The bug concerned the option to
compute random partitions of the landmarks into spatially
contiguous subsets. The algorithm for this particular option in
previous versions contained a bug that led to erroneous sets of
partitions and thus distributions of RV coefficients (for very
large numbers of random partitions, this distribution did not
necessarily approximate the distribution for the full enumeration
of partitions). In the present version, this algorithm has been
replaced by a simpler one that does not contain this error.
However, the simpler algorithm will check for many more partitions
of landmarks whether they are spatially contiguous or not. If
there are many subsets of landmarks or if the adjacency graph is
weakly connected (relatively few edges by comparison with the
number of landmarks), the algorithm can take a lot of computation
time. In this situation, consider to use a smaller number of
random partitions first to see how long it takes. Adding further
links to the adjacency graph will help to speed up computaions.
For the same procedure, a minor addition is a new item in the
pop-up menu for the diagrams with the distribution of RV
coefficients. This new command allows the user to change the size
of the arrow that indicates the RV coefficient for the hypothesis
of modularity.
A minor bug was fixed in the program that imports phylogenies
from Nexus files (Import Phylogeny
File). In the new version, all underscores in taxon and node
names are converted to spaces already in the initial import of the
Nexus file. This corresponds to the rule for Nexus files, and the
change therefore ensures that MorphoJ handles taxon names in the
way other programs do (or at least should).
Some further bugs related to phylogenies were also fixed,
affecting the permutation test and graphical output when the
symmetric component of centroid size (left-right averages for
structures with matching symmetry) is mapped onto a phylogeny.
This release includes a command for computing principal component scores using a dataset different from the one that may have been used to compute the principal components (PC Scores From Other PCA in the Variation menu). For instance, this can be used for between-group PCA, where the scores are computed for individual specimens, but the PCA is using the covariance matrix among the group means. It can also be used for phylogenetic PCA, where the PCA is run on the covariance matrix of independent contrasts (or a covariance matrix from a PGLS analysis that is imported into MorphoJ) and the scores are then computed for the terminal taxa of the phylogenetic tree. Other applications can be imagined too.
This version contains bug fixes for various analyses (CVA and others) using shape change vectors imported from outside MorphoJ (imported into the project by using Import Shape Change Vectors in the File menu). The new version also allows users to include or exclude observations from datasets containing shape change vectors. I came across these problems while doing analyses of data with complex symmetries.
The new release includes a fix for a particularly hard-to-find bug, which could produce crashes when saving a project, but only in rather particular circumstances. I finally found the bug after Hugo Benitez, Raffaella Bravi and the participants of the morphometrics course at Roma Tre university pointed it out to me and I was affected by it myself.
An additional minor bug fix concerns the display of graphs for matrix correlation if the graphs are invoked from a saved project (i.e., read from the prioject file rather than made from scratch in anew analysis). Thanks to Helmi Mohd Hadi Pritam for alerting me to the bug.
The release also contains a change in canonical variate analysis. Thanks to Benedikt HallgrÃmsson for pointing out the problem to me.
This release contains some bug fixes that caused failures when importing covariance matrices into MorphoJ. Thanks to Neus MartÃnez for alterting me to the bug.
In addition, the warning message upon exiting MorphoJ has been revised to indicate explicitly that any unsaved changes and analyses are lost. Thanks to Ruth Flatscher for pointing out the potential pitfall.
This release contains a bug fix concerning the computation of independent contrasts. For phylogenies that included branch lengths, earlier versions computed independent contrasts that used those branch lengths (i.e. with a weighted method) even if the option for weighting by branch length was not selected. The new version uses an unweighted method, where all branch lengths are set to the same lengths, if the option for weighting by branch lengths is deselected in the dialog box for Map Onto Phylogeny (in the Comparison menu). Thanks to Andrea Cardini for pointing out the bug.
This version adds a new functionality to the user interface for selecting a subset of landmarks (Preliminaries menu > Select Landmarks). The new user interface has "Move Up" and "Move Down" buttons that can be used to move landmarks or blocks of landmarks up and down in the list of landmarks that are to be included in the new dataset. This makes it possible to specify an order of landmarks for the new dataset that is different from the order in the old dataset. The new version also adds a function that can duplicate landmarks (a function that is expected to be used only rarely, if a landmark in one dataset corresponds to multiple landmarks in another dataset and the two datasets are to be combined in a joint analysis). Finally, the handling of paired landmarks has been improved in the new version, with an option that ensures automatically that paired landmarks are included or excluded as pairs.
The update also includes a couple of bug fixes in the user interface for extracting classifiers from the identifiers and for computing covariance matrices. For both, datasets that had no classifiers could produce crashes so that the respective dialog box failed to appear.
This version includes bug fixes in several places in the program that stop analyses if there are not enough degrees of freedom, and prevents crashes in those analyses or if the output is later used otherwise. It affects the following: Generate Covariance Matrix, Regression, Partial Least Squares (both between separate blocks and within a single configuarion of landmarks). Thanks to Soledad de Esteban and the participants of the morphometrics & phylogeny workshop in Sabadell for making me aware of the problem.
The program also fixes a bug in the computation of independent contrasts (when mapping data onto a phylogeny), which concerned the handling of zero-length branches. Thanks to Soledad de Esteban for telling me about the problem.
A further minor bug fix was added in the computation of PC scores from phylogenetically independent contrasts. For this kind of data, PC scores are now computed without centering (a contrast of zero will now produce a PC score of zero).
There are some major changes in how I go about programming and in the internal workings of this version. Above all, I now use version 6 of Java to write MorphoJ, and users therefore need to have a Java Runtime Environment of version 6 as well. This should not be a problem for Windows and Unix/Linux users, who can just download a new version from http://java.com for free. Mac OS X users may need to update their Java Runtime Environment (probably a good idea anyway, given recent security concerns). This change will enable me, in future versions, to add some new features that will make MorphoJ more convenient to use.
There is a bug fix in the method Modularity: Evaluate Hypothesis: the previous versions included an error in the computations of the RV coefficients for landmark configurations with object symmetry. Thanks to Emma Sherratt for pointing out the problem.
This version includes a new method: the command Compare Vector Directions in the Comparison menu peforms angular comparisons between vectors in shape space that result from multivariate analyses such as principal components, regression vectors, partial least-squares axes etc. The procedure provides the angles as well as P-values for the null hypothesis that the angles are no more similar than angles between random vectors drawn from a uniform distribution on a hypershere with the appropriate dimensionality.
The new version includes a bug fix for PLS analyses within a single configuration. For these analyses, if they were derived from a covariance matrix (rather than from a dataset), there was a bug in the routine reading information from a stored project file (with "Save Project" or "Save Projet As", and then "Open Project" or "Open Recent Project"). The bug affects the updating of the analyses and, for pooled within-group analyses, the displaying of the results. The bug us now fixed, but users should re-run within-configuration PLS analyses derived from covariance matrices and delete the old versions of those analyses from their projects (otherwise, the error persists in the old analyses). Thanks to Chuck Kimmel for alerting me to this bug!
The new version also can handle row and column labels in the files when importing covariance matrices.
Another minor bug was fixed, which produced a failure in drawing histograms if all values were the same (e.g. all zero). Thanks to José Tavera for reporting the problem.
This version contains a number of minor bug fixes, including problems with the user interface when switching outlines or wireframes in graphs of shape changes, and several changes to the user interface and to the computations of phylogenetically independent contrasts.
The most important change is that analyses involving independent
contrasts are now done without mean-centering (variances and
covariances are computed using deviations from the origin, not the
sample mean). This means, for instance, that regressions are
forced through the origin and covariance matrices are computed
from the deviations of independent contrasts around the origin.
This ensures that the order of taxa in the tree does not matter
(i.e., whether contrasts are computed as Left – Right sister node or Right – Left sister node). In practice, the
change from the previous versions that did not make this
distinction is usually small, because means of independent
contrasts are often close to zero (but this depends on the
ordering of taxa, and it is easy to set up a data set so that
there is a big difference!).
For this reason, some minor adjustments were made to principal
component analysis, partial least squares (both PLS between
separate blocks and PLS within a single configuration), and
regression analysis for handling independent contrasts.
The new version also contains a new feature that is likely to be helpful to users. Graphical output from MorphoJ is more flexible from this version on, because a new interface has been added, where a variety of options for the output of shape changes can be set (Set Options for Shape Graphs in the Preliminaries menu). The options include the color scheme, whether starting shapes and the numbers of landmarks are to be displayed, etc. These options are stored with the project and will therefore be retrieved if the project is saved and opened again later.
In addition, changes of the size of the points for landmarks in warped outline drawings and wireframe graphs (with "Resize Landmark Points" in the pop-up menu of the respective graph) are now stored with the outline or wireframe and will apply to all graphs generated with that outline or wireframe later (i.e., until it is changed again by the user). (For applying the change to the graphs already displayed in the Graphics tab, the user needs to generate a fresh copy using "Display Graphs" in the pop-up menu in the Project Tree window, or change the graph type at least once.) The changes will apply permanently to the respective wireframe or outline, that is, until point size is changed again, and are stored in the project file with the wireframe or outline.
Finally, if the user invokes the "Display Results" command in the pop-up menu of the Project Tree window for a dataset that has object symmetry, the pairing of landmarks is now included as part of the information that is displayed in the results window (previously, this was only in the output from a Procrustes fit).
This version contains a new method that addresses the problem of phylogenetic uncertainty in comparative analyses by collecting statistics from a set of trees that represent the range of uncertainty (Collect Statistics on Tree Set). For instance, phylogenetic uncertainty might be assessed by a bootstrap analysis; this method collects the tree statistics (tree length, P-value of permutation test for a phylogenetic signal) from all of the bootstrapped trees. The update also includes the associated import function for importing sets of phylogenetic trees to be processed together (Import Phylogeny Tree Set).
This version includes a bug fix for regression analysis, where some information was not passed along to the dataset of results, which might be necessary for some further analyses. Thanks to Eric Schmidt for pointing out the bug.
This version also simplifies the user interface for importing phylogeny files. It eliminates the option of attaching a phylogeny to a dataset in the project (this mostly caused trouble). Instead, all phylogenies are now attached at the base of the project tree.
This version corrects some errors in handling asymmetry of centroid size when mapping it onto a phylogeny. This concerns the permutation test as well as graphical and text output.
There are a couple of further bug fixes for PLS analysis within a landmark configuration. Above all, there was a bug in the code for the permutation tests that resulted in erroneous P-values. It was possible to select the option for pooled within-group analysis without specifying a classifier as the grouping criterion. Also, there is a change concerning the labelling of axes in scatter plots with superimposed phylogenies.
This version also contains a fix for a bug concerning PLS analysis between separate blocks. The permutation test for the correlations between PLS scores produced erroneous P-values (identical to those for the singular values -- the P-values are correct for those).
Thanks to the participants of the Sabadell workshop and Jana Makedonska for finding the bugs.
The new version contains a fix for one more bug that could result in a crash when mapping data onto unrooted phylogenetic trees and fixes for several minor bugs in the user interface (thanks to Paul Sanfilippo for pointing them out).
Various minor bug fixes include a problem that led to crashes when mapping traits onto unrooted trees and labelling of nodes (thanks to Ondřej Mikula for detecting the problems).
There are also some changes in the inner workings of this User's Guide, which fix a malfunction when viewing it off-line with Google Chrome or another version of the Chromium browser.
This version has a new analysis method: PLS analysis within a configuration of landmarks. This is useful for studying morphological integration in a single landmark configuration (e.g. among parts within a skull).
As another small addition, the user interface for mapping shapes onto phylogenies has an additional button with which the user can control whether the phylogeny is to be considered rooted or unrooted (so far, this was determined in the NEXUS file containing the phylogeny). For rooted trees, the location of the root is still determined in the NEXUS file (i.e. MorphoJ does not provide an option for re-rooting trees -- use software such as Mesquite for that purpose).
A number of minor bugs were found and fixed along with the work on the new method. They concern various aspects of the user interface, crashes when handling unrooted phylogenetic trees, the graphics for visualizing unrooted phylogenetic trees, pooled within-group covariance matrices and pooled within-group PLS analysis (with two separate blocks).
This version contains a new addition for scatter plots: an option for drawing confidence ellipses or equal frequency ellipses in the plots, either for all the data points together or for groups that are defined by a classifier. Thanks to Alexandra Wegmann for suggesting the addition.
This version contains some more minor bug fixes: importing NTSysPC files (thanks to Alan Robinson for reminding me of it) and another one that caused PLS analyses to crash under some circumstances (thanks to Anneke van Heteren for flagging it up).
It also has a new option for switching on and off labels with the identifiers for data points in scatter plots. Thanks to Viviana Toro for the suggestion.
This version adds an option to switch on or off the labels of taxa in phylogeny graphs. Switching them off is convenient for studies where there are many taxa. Thanks to Emma Sherratt and Yoshi Yonehara for suggesting this option.
In addition, a bunch of bugs concerning the user interface and updating of projects have been identified and fixed. Thanks to Yoshi Yonehara for bringing them to my attention.
This version contains a number of bug fixes: for exporting coordinate changes from 3D graphs for visualization in Landmark on machines using non-English operating systems, for handling of missing observations in regression. Thanks to David Galicia for demonstrating the problem.
As a new addition, this version has an option to export covariance matrices, in the File menu.
This version adds a scatter plot of group-centered regression scores to the graphical output for pooled within-group regression.
This version contains some improvements of small annoying things in the user interface.
A bigger change is a fix of a bug in the calculation of sums of squares for pooled within-group regression. The revised version computes the within-group sums of squares (predicted, residual and total). The percentage of the variation accounted for therefore also refers to the pooled within-group variation. This is the correct method because this type of regression analysis concerns within-group variation and should be evaluated accordingly. Note that the residuals and predicted values in the results dataset are computed from the total deviations from the grand mean (i.e. the deviations of the group means from the grand mean are also divided into predicted and residual components).
This version fixes some minor bugs concerning user interaction in the context of defining wireframes and generating pooled within-group covariance matrices, as well as a bug concerning the handling of excluded observations in two-block partial least squares. Thanks to Anneke van Heteren for reporting the latter one.
This version fixes a minor bug: contrary to what is stated in the User's Guide, data could not be imported from text files with column titles in the first line. This is now possible. Thanks to Elferra Swart and Emma Sherratt for reporting the problem.
2-block partial least squares: There was a bug in the computation of the permutation test for pooled within-group PLS analyses. Effectively, the output in earlier versions was the P-values for a PLS analysis not taking into account the group structure. The bug has been fixed.
Thanks to Chuck Kimmel and Benedikt HallgrÃmsson for spotting the bug!
2-block partial least squares: The main change in the new version is a bug fix in the computations of the P-values for the permutation tests of the correlations between corresponding PLS scores. The test for these correlations in the previous versions was erroneous. But even after the bug is fixed, you should be careful in interpreting the results of this test, and the correlation between the corresponding PLS scores in the two blocks is probably best taken as a descriptive statistic (i.e., best ignore the P-value). The problem is that these correlations can behave strangely if sample sizes are small or if there are dimensions with very little or no variation. Thanks to Andrea Cardini for flagging this up.
The new version also contains fixes for a variety of minor bugs (unhelpful error messages etc.). Thanks to all who reported these occurrences.
The new version adds the possibility to color the dots representing the terminal taxa in phylogeny plots according to the values of a classifier (e.g. ecological or biogeographical information). Only those classifiers are considered that have identical values within all of the taxa. Thanks to Emma Sherratt and the participants of the Barcelona morphometrics/phylogeny workshop for the request.
The new version also made a change in scatter plots for which data points are colored according to classifiers: the legend lists the groups in alphabetical order.
Average observations by...: Fixed a bug that resulted in some data types being omitted from the resulting dataset. Thanks to Tom Ellis for pointing this out.
Phylogeny: Fixed bugs in the datasets with branch differences and independent contrasts, which prevented them to be used in further analyses like CVA. Thanks to the participants of the Barcelona morphometrics/phylogeny workshop for finding the bug.
Modularity: Evaluate Hypothesis: A bug in the interface for editing adjacency graphs was fixed: for configurations with object symmetry, corresponding links are now added or deleted on both sides simultaneously. Thanks to Emma Sherratt for alerting me to the bug.
The new version fixes a problem in the phylogeny procedure with
missing values for covariates. If there were missing data in the
covariates, the previous versions produced no results or crashed.
This has now been fixed.
For discriminant function analyses, the new version now produces
an output dataset with the discriminant scores and the scores from
the cross-validation procedure. This dataset can be used to check
how particular specimens were classified.
Thanks to Emma Sherratt and Kim Ballare for pointing out those
things.
Above all, this version contains the fix for the bug in the
computation of residuals from pooled within-group regression.
Values computed with older versions were erroneous.
All users who have used residuals from pooled within-group
regressions in their work should download the new version and
repeat the regression analyses (and discard the old analyses; in
the Project Tree window, using 'Remove Item' in the pop-up menu of
those regression analyses). I sincerely apologize for the
inconvenience that this may cause you.
The new version also contains a new method, Residuals/Predicted Values
From Other Regression in the Covariation menu of the
program. This enables the user to use a regression to compute
residuals and predicted values for a dataset different from the
one used to set up the regression.
For instance, a user might be interested in the role of
ontogenetic scaling for the diversification in a group of
organisms. To characterize ontogenetic allometry, a regression of
shape on centroid size is done for data of growth series in one or
more species (if there are several species, this would be done by
a pooled within-species regression). To look at the
diversification in the whole group, this regression would then be
applied to a dataset of adult variation in all species of the
group.
I'd like to thank the many people who have made comments and
suggestions on the features in MorphoJ. For the current series of
updates, Emma Sherratt and Andrea Cardini were particularly
helpful.
Various minor bug fixes and enhancements. For instance, it is possible to create a new dataset from multiple input files (text, NTSys, etc.), as long as the dimensionality and number of landmarks agree. A number of other changes removed unnecessary restrictions for the choice of data in particular analyses, etc.
Moreover, there is an additional possibility for visualizing shape changes in 3D data with the Landmark software, based on information exported from MorphoJ. Click this link for more details.
Added the new method Modularity: Evaluate Hypothesis.
Various bug fixes, including one that restored the SVG file functionality before version 1.00k.
First, I have made some changes to the user interface
(particularly to the arrangement of items in the File menu) to
bring it better into alignment with general standards for user
interfaces (thanks to Mauro Cavalcanti for pointing me to this
issue).
Second, I have replaced the library used to export graphs. MorphoJ
now can export graphs in a range of different file formats:
Vector graphics formats:
o *.emf, Windows vector graphics
o *.eps, Encapsulated PostScript
o *.ps, PostScript
o *.pdf, Portable Document Format
o *.svg, Scalable Vector Graphics
o *.swf, Macromedia Flash
Raster graphics formats:
o *.gif, Graphics Interchange Format
o *.bmp, bitmap image format
o *.png, Portable Network Graphics
The vector graphics formats are preferable because they allow
editing of the parts of the graph and can be displayed at any
magnification without loss of resolution.
Finally, a number of bugs have been fixed. These affected a number
of 'housekeeping' functions, Procrustes ANOVAs (with some
combinations of effects included in the ANOVA model, there were
errors in the F-statistics and associated P-values),
and subdividing datasets if there are missing values in the
dataset (thanks to Nandini Singh for reporting that problem).
The first, which went into version 1.00i, is a bug fix that
corrects a mistake in the calculation of the P-value for
the T-square test that is computed with Discriminant
Function analyses. The faulty calculation produced P-values
that were too large (overly conservative) for relatively small
samples.
Thanks to Andrea Cardini for spotting the problem.
The second update (in version 1.00j) concerns the inner workings
of the program, how project files are read. The fix eliminates (I
hope!) the problem that the project tree for larger projects was
sometimes messed up. It will also make it simpler for me to add
further types of analyses. There should be no difference in the
functioning of the program between versions 1.00i and 1.00j.
Moreover, Mauro Marcinkevicius alerted me to the fact that the
larger installer files, as they are distributed since version
1.00h (i.e., including a copy of the documentation), can produce
problems because some internet providers appear to enforce a limit
on file sizes. Accordingly, I have started distributing a web
installer for each of the  suported platforms (Windows, Mac,
Linux/Unix). The web installer will download the files necessary
for the installation in three separate chunks. The downloaded
files are therefore smaller, but the web installer can only be
used for computers that are connected to the web at the time of
installation.
This update fixes a minor bug concerning the editing of
covariates (there was a problem with adding covariates to a
dataset that did not have any before).
The main change, however, is the addition of a Help menu with a command User's Guide.
This launches the default browser on the user's machine to display
the MorphoJ user's guide, which is now distributed automatically
with the installation. Therefore, the user's guide is now also
available for off-line use.
As a result, the installer files for all operating systems have
become a bit bigger.
Thanks to Andrea Cardini for the impetus to look into this.
Version 1.00f, posted a couple of weeks ago, added a new command
Open Recent Project in the
File menu. MorphoJ keeps a record of the 10 project files that
were opened or saved most recently. These files are available for
opening without having to select the entire file path.
Thanks to Chris Harrod for the suggestion.
Version 1.00g provides an improvement of the user interface for
swapping pairs of landmarks that is available under Find Outliers in the
Preliminaries menu. If more than one pair of landmarks need to be
swapped for a configuration, the user can make all the necessary
swaps (and see the changes) until the 'Done' button is pressed to
return to the main interface.
Version 1.00d was posted a while ago. It included an improved
version of the algorithm used to guess the paired and unpaired
landmarks for landmark configurations with object symmetry. The
revised algorithm makes better use of the available information
and should therefore require less manual intervention by the user
as part of the procedure for the Procrustes fit, even if the
configuration is round overall and has many, densely spaced
landmarks (e.g. human skulls). Thanks to Andrea Cardini for
pointing out the problem.
Version 1.00e contains a few bug fixes (e.g. keeping track
correctly of the wireframes and outline drawings in PLS analyses
with two landmark configurations; renaming data matrices).
But above all, the new version adds the option of running a pooled
within-group version of two-block PLS analysis. If there are
multiple groups (e.g. species) in a study, this option allows to
do a PLS analysis of the covariation between the deviations from
the group averages in the two blocks of variables.
Thanks to Nandini Singh for alerting me to the problems and
convincing me that the new PLS option should be included.
There was a problem with processing the row labels from .NTS
files and a bug concerning Morphologika files that did not have a
[names] clause (thanks to Andrea Cardini for alerting me to these
problems).
Moreover, there was a bug that made it impossible to save projects
containing discriminant function analyses without the permutation
test.
The update rectifies a couple of bugs in the Procrustes ANOVA.
The first bug concerned the output only, but not the results of
the analysis. If a 'Side' effect is specified or if the data have
object symmetry, the vector of directional asymmetry (average
asymmetry) is now printed to the Results window.
The second bug concerned the covariance matrix for the Individual
effect that is produced as an output of the Procrustes ANOVA,
under particular circumstances (no Error 1 or Error 2 effect, but
multiple repeat measurements included as residual effect). The
problem was that divisions by zero could occur in the computations
of covariances, which led to failures of subsequent analyses such
as PCA, matrix correlation etc. These problems were obvious (e.g.
an output "NaN" for the matrix correlation or complete failure of
the procedure for PCA).
If you have any Procrustes ANOVAs and want to abe absolutely safe,
remove the Procrustes ANOVAs from your existing projects (invoke
the popup menu in the project tree window, then select 'Remove
Item') and then redo them.
This was the original release of MorphoJ.