Diversity Dependent Diversification

Description

This package is a likelihood-based statistical package to estimate parameters under diversity-dependent diversification models.

New in version 0.3:
- Models with shifts in the parameters at a time tshift.

New in version 0.4:
- Bug fix of prefactor in likelihood - bug did not affect parameter estimates.
- There is an option to condition or not on non-extinction of the phylogeny.
- The functions dd_loglik and dd_SR_loglik now accept untransformed parameters; the transformation is done in the auxiliary functions.
- Improved help pages.
- Auxiliary functions no longer in help pages.

New in version 1.0:
- Another bug fix of prefactor in likelihood - bug did not affect parameter estimates.
- Models with key innovations defined as decoupling of diversity-dependent dynamics.
- Option to compute the likelihood of a set of branching times or of the phylogeny; these differ by a constant factor.

New in version 1.1:
- A model with diversity-dependence in speciation and extinction rate, currently only available for the likelihood without shifts or decoupling.

New in version 1.2:
- Possibility of automatically exploring multiple initial conditons for the SR models
- Suppressing unnecessary output
- Flushing computed likelihood values immediately (even with buffer on)
- Bug fix in default value of the initial time of decoupling

New in version 1.3:
- Fix of default length of idparsopt
- Fix of conversion problems for small branching times
- Default initial parameter values in dd_ML based on birth-death model

New in version 1.4:
- Bug fix for when there is a large number of missing species

New in version 1.5:
- Bug fix for returned output of functions when there are fixed parameters.
Textual output of DDD was correct.
- Return to Nelder-Mead optimization for dd_ML (was set to "nlm" in versions 1.3 & 1.4)
- New format for output off all ML-functions (dd_ML, dd_SR_ML, dd_KI_ML)

New in version 1.6 (not on CRAN)
- Bug fix in returned result when optimization does not converge.
- Works also when only branching times are entered in ascending order (previously error was reported)

New in version 1.7
- Cleaning up code, fixing various small bugs
- New optimization algorithm because of occasional strange results with optim package

New in version 1.8
- Bug fix for missing number of species in KI model
- Bug fix for optimization of one parameter (due to R's inconsistent handling of matrices)
- Model 5 (dd_ML and dd_loglik) now also allows values of -1 < r < 0; i.e. when extinction also decreases with diversity but less so than speciation

New in version 1.9
- Bug fix in optimizer (did not stop when likelihood was positive)
- Bug fix for large numbers of missing species
- Conv = -1 is now always reported when optimization is not started or does not converge
- An error is reported when parameter values cause numerical problems (NA or NaN) in likelihood computation, e.g. when parameter values are very high; loglik is then set to -Inf
- For dd_KI_loglik and dd_KI_ML: if the number of missing species in main clade and subclade is known, this can now be specified
- Fixing K at Inf is now possible in SR and KI models

New in version 1.10
- All non-convergences now get conv = -1 (this was conv = 1 for true nonconvergence and conv = -1 for optimization that did not start because of low likelihood values)
- Likelihoods can also be computed for a tree with a stem age rather than only a crown age

New in version 1.11
- Various small tweaks to make the code run faster and some bug fixes
New in version 1.12
- Allows conditioning on survival of crown linages AND on the number of extant species by setting cond = 2

New in version 1.13
- Fast computation of dd_loglik for the case K = Inf

New in version 1.14
- Fast computation of dd_loglik for the case mu = 0
- Allows time-dependence instead of diversity-dependence; exponential decline in speciation rate and/or extinction
- Start of time-dependence can be specified to allow synchronisation across multiple clades
- Removed link to LASER

New in version 2.0
- Includes function to simulate trees under diversity-dependence

New in version 2.1
- Fixes numerical problems with the convolve-function (used to deal with missing species)
- Fixes a bug in dd_KI_loglik (ddep unknown)
- Changed third element of the output of dd_sim, such that all speciation and extinction times are in Mya

New in version 2.2
- Allows conditioning on the number of taxa only for dd_loglik, asssuming a uniform prior on stem age; this is now the default
- An additional model of time-dependence, following deterministic logistic equation

New in version 2.3
- Allows conditioning on the number of taxa only for dd_SR_loglik, assuming a uniform prior on stem age; this is now the default
- Fix of bug for stem age (soc = 1), introduced in v2.2.
- Allows Rabosky's exponential model, with la(n) = la_0 * n^(-x(la_0,mu,K)) as ddmodel = 2.1, mu(n) = mu_0 * n^x(la, mu_0, K) as ddmodel = 4.1
- Allows specific case of Rabosky's exponential model, with la(n) = la_0 * n^(-1), as ddmodel = 2.2, and mu(n) = mu_0 * n, as ddmodel = 4.2

New in version 2.4
- Fix of bug in dd_KI_loglik ('m' was not found)

New in version 2.5
- Fix of bug in bd_loglik where NA was output to screen for a nonexistent parameter under diversity-dependence without extinction
- Fix of bug in dd_KI_loglik when m is a two-element vector
- Fix of bug in dd_KI_loglik, introduced in v2.3

New in version 2.6
- Allows linear dependence of speciation with parameter K' (= diversity where speciation = 0)

New in version 2.7
- Fixes bugs in dd_KI_loglik, for cond = 0 introduced in v2.3, and for cond = 1 introduced in v2.0
- Solves (some) memory problems with dd_KI_ML and dd_KI_loglik when cond = 1
- Provides option to set loglikelihood to -Inf if optimization did not converge

New in version 3.0
- Includes a model of macroevolutionary succession in which there is no decoupling of diversity-dependence but parameters may change
- Computes LRb, a bootstrap version of the likelihood ratio
- Allows use of alternative ODE solvers
- Allows use of subplex as optimization algorithm.

New in version 3.1
- Fixes bug with dd_sim and dd_KI_sim for ddmodel 1.3 (Thanks to Giovanni Laudanno)
- Fixes bug with cond = 3
- Includes matrix exponentiation as an alternative to numerical ODE solvers

New in version 3.2
- Fixes bug with dd_loglik when lambda becomes infinite
- Changes default tolerance of subplex
- Fixes tiny bug in p-value computation in dd_LR

Forthcoming:
- High-precision arithmetic in numerically unstable regions of bd_loglik

Details

bd_loglik compues the likelihood under the constant-rate birth-death model of several specific time-dependent birth-death models.

bd_ML finds the parameters that maximizes the likelihood computed by bd_loglik.

dd_loglik computes the likelihood of any of 4 different models of diversity-dependent diversification, given a set of parameters and a data set of phylogenetic branching times.

dd_ML finds the parameters that maximizes the likelihood computed by dd_loglik.

dd_SR_loglik is the same as dd_loglik except that it allows a shift in the parameters at time tshift.

dd_SR_ML finds the parameters that maximizes the likelihood computed by dd_SR_loglik.

dd_KI_loglik is the same as dd_loglik except that it allows decoupling of a subclade's dynamics from the main clade's dynamics potentially accompanied by a shift in parameters at time t_d.

dd_KI_ML finds the parameters that maximizes the likelihood computed by dd_KI_loglik.

dd_MS_loglik is the same as dd_KI_loglik except that the subclade's dynamics do not decouple from the main clade's dynamics but they do get different parameters at time t_d.

dd_MS_ML finds the parameters that maximizes the likelihood computed by dd_MS_loglik.

dd_sim simulates the diversity-dependent process

dd_LR performs a bootstrap likelihood ratio test

Author(s)

Rampal S. Etienne & Bart Haegeman
Maintainer: Rampal S. Etienne (r.s.etienne@rug.nl)

References

- Etienne, R.S. et al. 2012, Proc. Roy. Soc. B 279: 1300-1309, doi: 10.1098/rspb.2011.1439
- Etienne, R.S. & B. Haegeman 2012. Am. Nat. 180: E75-E89, doi: 10.1086/667574

See Also

TESS

Results