diversitree: Changes and releases
- 7 May 2012 [0.9-3] Mostly
a bug fix release, but includes features:
-
A much faster BM/OU likelihood calculator
method="pruning", backend="C"
, and
optimisations for the method="vcv"
BM
calculator (also note that the "direct" backend is renamed
"pruning").
-
GeoSSE tree simuator now included
(
tree.geosse
and via trees
))
-
sim.character
now faster.
-
BiSSE, MuSSE/multitrait and Mkn/multitrait allow logical
trait values, in addition to binary (0=FALSE, 1=TRUE).
-
trait.plot
automatically collapses
non-monophyletic groups.
Notable bug fixes:
-
make.musse.multitrait()
renamed traits when
called with single trait.
-
make.prior.uniform
used wrong value for
out-of-bounds parameters.
-
BM and OU were mislabelled in
sim.character
.
- 25 March 2012 [0.9-2]
New models and extensions to models:
- geiger-style vcv based calculations for
make.ou
, via method="vcv"
.
Bug fixes, etc:
sim.character
was broken in 0.9-1
Improved profiles.plot
(automatic
legends, better colour control).
- Improved likelihood function printing.
- This version also has had an enormous behind-the-scenes
rewrite (again). Despite the small list of changes,
thousands of lines of code have changed, and bugs may still
remain.
- 13 March 2012 [0.9-1] New models and
extensions to models:
-
Includes BiSSE-ness (by Karen Magnuson-Ford and Sally Otto) and ClaSSE (by Emma Goldberg),
extending BiSSE and MuSSE models to non-anagenetic
evolution.
-
make.geosse.split
(again, contributed by Emma Goldberg) for
incorporating topological rate varaition in GeoSSE models.
-
Added
make.mkn.multitrait
for modelling
multiple binary traits (analagous to
make.musse.multitrait
).
-
Alternate method of calculation for Mk models
(
make.mkn
) with very large numbers of
possible states. Select with
control=list(method="ode").
-
Non-VCV based Brownian motion likelihood calculations (via
the argument
control=list(method="direct")
to
make.bm
). This can be substantially faster
for very large trees (> 1000 tips).
-
New version of the
mkn
model that will be
substantially faster for many traits (> 30 traits).
This mainly affects mkn.multitrait
.
-
Ornstein-Uhlenbeck model (for completeness, as this is
present in many,
many
different
packages
on CRAN already).
-
Function
sim.character
for simulating traits
given a tree.
-
Ancestral state reconstruction under split BiSSE and MuSSE
models.
Notable new features and bug fixes:
-
Updated to work with deSolve 1.10-3
-
mcmc
can now return partial results during a
long run.
-
make.prior.uniform
for specifying uniform
priors.
-
make.bd.t
did not allow incomplete sampling
(reported by Dan Rabosky).
-
Improved the way
contrain
deals with
constraining parameters that are targets of constraints.
Also, additional arguments were not being passed to the
functions generated by constrain
.
-
"Split" functions now allow caching branches (via
control=list(caching.branches=TRUE)
). This
can result in large speedups, particularly for
make.quasse.split
.
-
Improved consistency of calculation method switching
across different likelihood functions (especially
bd
and bm
models).
-
Simulate trees under the
musse.multitrait
model.
-
The
anova
function will now refuse to run a
likelihood ratio test for non-nested models (it may be
overzealous in some cases).
-
Function
trait.plot
to plot the distribution
of multiple traits on a radial phylogeny.
- 18 July 2011 [0.7-7]
- Works with deSolve 1.10-2.
- 11 July 2011 [0.7-6]
- New parametrisation of MuSSE available where the
multi-state trait is really combinations of binary states.
This allows expressing speciation and extinction rates as main
effects and interations of the underlying binary traits, and
easily restricting trait evolution to a single trait at a
time.
- Some internal streamlining that should speed up most
models slightly
- Faster stochastic character mapping
- Various bugs fixed (
mcmc
not passing
additional arguments properly, make.bd.split
not
using unresolved clade information).
- 12 May 2011 [0.7-2]
- Fixed bug that prevented use on Windows
- 11 May 2011 [0.7-1]
- New Sundials
based solver CVODE
available (switch with
control=list(backend="CVODES")
). In some cases this is
8x faster than the previous version.
- Bug fixes. In particular, the equilibrium root state
treatment (
ROOT.EQUI
) was incorrect.
- Works with deSolve 1.10.
- 20 April 2011 [0.6-5]
Bug fixes mostly:
- Better integration control for all ODE-based solvers.
- Slightly altered root state treatment (exact calculations
will change with
condition.surv=TRUE
)
- Very basic mixed integer nonlinear optimisation approach
for finding ML point for some QuaSSE models that involve a
character space variable (poorly documented at present).
- Assorted minor bug fixes.
- 26 January 2011 [0.6-3]. Bug
fixes and some new features/models:
- Includes Emma
Goldberg's GeoSSE model
- New numerical method-of-lines
integrator for QuaSSE, which may be more robust in some
situations.
- Marginal ancestral state reconstructions for MuSSE
models.
- Better support for MuSSE/mkn models with >9 states
- Works with deSolve 1.9
- Builds on CRAN
for windows!
- 21 October 2010 [0.6-1]. A
number of new models.
- "QuaSSE" (Quantitative Trait Speciation and
Extinction): new method is now released. This tests for
correlation between a continuous/quantitative character and
rates of speciation and/or extinction. Includes support for
"split" QuaSSE analyses.
- Improved support for time dependent models: now,
speciation, extinction, and (for BiSSE only) character change
rates may be arbitrary functions of time. This is implemented
for birth-death models, BiSSE, and MuSSE.
- 23 September 2010 [0.5-3]. A couple
of model extensions, documentation improvements, and bug fixes.
- Now works with deSolve 1.8
- Time-dependent and "split" MuSSE models are now included.
The interface for both is similar to their BiSSE
analogues.
- A new tutorial document.
- 12 August 2010 [0.5-2]. A
number of new models and bug fixes.
- "MuSSE" (Multi-State Speciation and Extinction):
the multiple-character extension to BiSSE is now implemented.
This allows modelling characters with more than two states, or
combinations of characters (similar to the Pagel 1994 method
for rates of character evolution). Likelihood calculations
and tree simulation are supported.
- Simple support for Brownian Motion models of continuous
trait evolution (single trait only).
- "Split BiSSE" model: this combines aspects of the MEDUSA
model (Alfaro et al. 2009) with the BiSSE model to allow
different parts of the tree to have different rates of
speciation, extinction, and/or character evolution. For
completeness a "Split BD" model is also implemented. None of
these assist in searching for positions in the tree where
rates shift.
- "Time-dependent" BiSSE, where different chunks of time
have different rates of speciation, extinction, and/or
character evolution.
The new models' documentation is still very preliminary and
these are not yet in the tutorial document.
Other new features:
- Interrupting a
mcmc
run will return a
truncated set of samples.
- Constrained functions can be constrained (though this is
still limited)
subplex
is now the default ML search
algorithm for BiSSE and MuSSE models.
- 2 March 2010 [0.4-5]. Some
important bug fixes (BiSSE likelihood calculations were very
slightly incorrect at root and unresolved clade calculations
were incorrect when no species within a clade had a known
state). Other changes:
- Added support for sampling and unresolved clades in
constant rate birth-death models.
- Fixes to the Mk2 ancestral state reconstructions
- Works with deSolve version 1.7
- Assorted small bug fixes and documentation
improvements
- 2 Febrary 2010 [0.4-4]. New
features include:
- Ancestral state reconstruction and stochastic character
mapping under Mk2 (
?asr.mk2
) (slow, but will get
faster in future versions)
- Much faster Mk2 likelihood calculation (speed increase of
20-50x)
- Ancestral state reconstruction under BiSSE
(
?asr.marginal.bisse
)
The ancestral state reconstruction code should be regarded as
experimental.
- 19 January 2010 [0.4-3]. Mostly
a few bug fixes.
- 11 January 2010 [0.4-2]. New
features include:
- Remembering and plotting character histories on simulated
trees (see
?tree.bisse
- Works with deSolve 1.6
- Support for simulating Yule trees and estimating Yule
model parameters
- Bug fixes and improvements to the MLE code
- 02 December 2009 [0.4-1]. New
features include
- Constant rate birth death model (
make.bd
) and
Mk2/Mk-n (make.mk2
, make.mkn
) models
implemented
- Now works with the current deSolve package
- Node states now retained during tree simulation (see
?trees
for an example)
- Simple MCMC plots with
profiles.plot
- Several bugs fixes, documentation improvements, additional
examples, and minor features
- 16 October 2009 [0.3-2]:: Minor bugfix
- 15 October 2009 [0.3-1]:
- New formula-based constraint interface (through function
constrain()
)
- Subplex available to do ML searches
- Simple hypothesis testing convenience functions added
(
anova()
, logLik()
).
- 5 May 2009 [0.2-1]: Initial public release