Rev Language Reference

• dnPhyloBMSD
• dnPhBMSD

dnPhyloBrownianProcessStateDependent(CharacterHistory characterHistory, RealPos sigma, Real rootValue, Natural nSites)

characterHistory :	CharacterHistory (pass by const reference)
	The character history object from which we obtain the state indices.
sigma :	RealPos (pass by const reference)
	The rate of random drift (per state).
	Default : 1
rootValue :	Real (pass by const reference)
	The value of the continuous trait at the root.
nSites :	Natural (pass by value)
	The number of sites which is used for the initialized (random draw) from this distribution.
	Default : 1

ContinuousCharacterData

The phylogenetic state-dependent Brownian process is collapsed from the phylogenetic state-dependent Ornstein-Uhlenbeck process, in which `alpha` is always 0. The probability is computed using a pruning algorithm. Specifically, `dnPhyloBMSD` uses an augmented tree structure, i.e., the character history with different character states (also referred to as "regimes"; Hansen 1997). Shifts in the character state are discrete, do not occur concurrently with speciation events, and can be represented by nodes of degree 2. As such, each branch in the augmented tree structure takes exactly one character state. Under this process, branches of the same character state share the same rate parameter. These character states can correspond to the states of an observed discrete character. This model is equivalent to the MuSSCRat model (May and Moore 2020) if (1) the continuous trait is univariate, and (2) we assume no background rate variation. The major difference between these two models lies in the probability computation. The MuSSCRat model transforms the branch lengths according to the `sigma` parameter and computes the probability using a state-independent algorithm. The phyloBMSD model does not transform the phylogeny and computes the probability using a state-dependent algorithm. Applications of this model include: 1. State-dependent continuous trait evolution conditional on discrete character history In this application, the discrete character history (in `simmap` format) is read (using `readCharacterHistory`) and specified in the `characterHistory` argument. The `sigma` arugments represents the state-dependent diffusion parameter that controls continuous trait evolution in different discrete character states. The continuous trait observations are fixed to the tips using `.clamp()`. 2. Joint inference of discrete character history and state-dependent continuous trait evolution In this application, the discrete character evolution is modeled using the distribution `dnPhyloCTMCDASiteIID` (see Landis et al. 2013, May and Moore 2020). The character history is retrieved from the distribution (using `.characterHistories`) and specified in the `characterHistory` argument. Other parameter specifications are same as above.

# setup for a two-state phyloOUSD model
num_states = 2          # 0 and 1 are the only states

# option 1: conditioning on a fixed character history
char_hist = readCharacterHistory( simmap_path )[1]

# option 2: joint inference of character history
Q <- fnJC(num_disc_states)
X ~ dnPhyloCTMCDASiteIID(tree, Q, branchRates=1, type="Standard")

# set state-dependent OU parameters
for (i in 1:num_states){
  sigma2[i] ~ dnLognormal(1, 0.587405)
}

# basic use of the function
Y ~ dnPhyloBMSD(char_hist, sigma=sigma2^0.5, rootValue=2.5)

• dnPhyloOUSD
• dnPhyloCTMCDASiteIID
• readCharacterHistory