Dear Colleagues,

We wish to alert you to major updates in geomorph and RRPP, now available on 
CRAN.  These updates should not introduce any operational differences in your 
analyses, but will give you various additional options, plus provide much 
greater computational efficiency, especially if you have large data sets.  Here 
is an overview of the improvements:

1. The gpagen function in geomorph v4.0.0 now has the following additional 
arguments (options) and features:
Parallel: a logical or numeric argument to allow parallel processing during GPA 
and define how many cores to use.  Parallel processing is available on all 
operating systems.
appBE: an option to use bending energy estimated from approximated thin-plate 
spline for sliding semi-landmarks.  This argument, plus its associated 
sensitivity, allows one to use a subsample of landmarks rather than the whole 
set for estimating bending energy, which can greatly reduce the size of the 
bending energy matrix and make computations much faster.  This can also mean 
increasing GPA iterations to find better convergence.
Sparse linear algebra: gpagen algorithms now use sparse linear algebra, which 
reduces computer memory demands and allows for faster computations.
max.iter: The default for the maximum number of iterations is now 10 instead of 
5, as a result of computational improvements.  
verbose: A logical argument to suppress various GPA statistics (like Procrustes 
distance matrix or landmark variances), which can require a lot of computer 
memory for large data sets.
Collectively, these updates make gpagen a much faster and computer 
memory-efficient function.  For most users, gpagen should seem the same but for 
those with large data sets, the improvements could be profound.

2. ANOVA functions (procD.lm and bilat.symmetry functions in geomorph v4.0.0; 
lm.rrpp in RRPP v1.0.0) now feature the following enhancements:
Algorithms to choose among several options for estimating sums of squares, 
based on which is most computationally efficient, both in terms of speed and 
computer memory.
Better parallelization options, including parallel processing for Windows PC 
operating systems
A new “turbo” argument, which suppresses coefficient estimation in every RRPP 
permutation, if not needed, making analyses much faster.
Likewise, these updates make ANOVA-producing functions much faster and more 
efficient in terms of computer memory.  For users with large data sets, the 
same results should be found as before the updates, but now in much shorter 
time.

3. The convert2ggplot function in RRPP v1.0.0 and the make_ggplot function in 
geomorph v4.0.0 are new functions to produce ggplot facsimiles of these 
packages’ plots, which are subsequently amendable. These functions might be 
useful for users who prefer working with ggplot objects.  These functions 
coerce plot parameters into ggplot parameters and are intended as courtesies to 
quickly generate ggplot templates.  Users with strong plotting skills will 
probably prefer to still work from the ground up making plots “by hand”.

We hope these enhancements will be helpful.  Any issues with the new packages 
can be reported at [email protected] 
<mailto:[email protected]>.  Please note that RRPP v1.0.0 
must be installed in order to run geomorph v 4.0.0.  Both packages are 
available on CRAN and updated regularly on github (see Description files).

Thanks and happy computing!

Mike, and the geomorph team

-- 
You received this message because you are subscribed to the Google Groups 
"Morphmet" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to [email protected].
To view this discussion on the web visit 
https://groups.google.com/d/msgid/morphmet2/84B9BD4F-886C-4DAE-9AC4-E31C5B1C7298%40gmail.com.

Reply via email to