r8s Files

** r8s v. 1.81; May 2016 **

This release of r8s is unfortunately a slightly retrogressive version.
Some features have been removed as they relied on code from "Numerical
Recipes in C." The core functionality of inferring divergence times
using Langley-Fitch (clock-based maximum likelihood) and penalized 
likelihood has been retained. The optimization algorithm that is
retained for both these methods is the truncated Newton ("TN") algorithm.
The implementation for TN is fortran code kindly left in the public
domain by Stephen Nash, George Mason University (see the source code).
Other methods and algorithms are no longer supported (at least not
at the moment), but I have found TN to be the most robust algorithm
in most data sets anyway. TN requires analytical gradients, which can
be calculated for LF and PL, but these are not yet done for local
clock methods, for precursor inference. Stay tuned for those. There 
may well be other instances where the code does not work. Thanks
to Andreas Tille for bringing licensing issues with NRC to my
attention. 

INSTALLATION: See the manual, which contains installation procedures
for Mac OS X and Linux. Makefiles for both are contained
in the src directory. I have also included a binary compiled for
OS X. The manual is a bit dated, but in most cases typing

	make

in the src directory on OS X should be enough, or

	make -f Makefile.linux

on a linux machine.


EXAMPLES: The examples directory contains some examples that work,
though not all do now. From the src directory, after compilation, try

	./r8s -f ../examples/SAMPLE_SIMPLE

which should work.

LICENSE: The C code written by M.J. Sanderson is distributed under the
GPL (see included copy). Fortran code by Stephen Nash is not included
in this license. See that source code for any questions.