make_BayesCMR: Function to construct one clade CMR-model.

In josteist/Compadre: Analysis of Macroevolutionary Rates of Speciation, Extinction and Sampling using Capture-Mark-Recapture techniques

Description

This function generates a structure containing the necessary functions for a CMR analysis of a fossil dataset.

Usage

make_BayesCMR(
  Obs,
  dts = rep(1, dim(Obs)[2]),
  spec = ~1,
  ext = ~1,
  samp = ~1,
  data = NULL,
  pfix = 2,
  priorMu = dnorm,
  priorCov = dnorm,
  priorStd = dunif,
  priorPars = list(list(log(0.1), 5), list(0, 5), list(-3, 3)),
  modeltype = "III"
)

Arguments

Obs	a matrix with size number of taxa by number of intervals (n). Each taxon has a row with 0's (unobserved) and 1's (observed) for each interval in the analysis. Oldest interval is first column.
dts	a vector of interval durations. Defaults to a series of 1's if not supplied. Oldest interval is first entry.
data	a data frame with putative drivers.
pfix	a switch to select which approach is used to solve the identifiability problem in the model. If pfix = 1, then the sampling rate in the first and the last intervals are assumed to be equal to the mean sampling rate for the whole period. If pfix = 2, then the first two intervals, and the last two intervals have the same sampling rate. Defaults to pfix = 2
priorPars	list of lists of the parameters for the priors. First list is parameters for baseline rates, then list of parameters for covariates and third is list of parameters for ln of standard deviation of random effects))). Defaults to list(list(log(0.1),5),list(0,5),list(-3,3)).
modeltype	sets the type parameterisation of the rates of speciation and extinction. 'I' and 'II' specifies that the probabilities of extinction and seniority are calculated based on actual durations of intervals, whereas 'III' and 'IV' uses the differences between mid-points of intervals. 'I' and 'III' uses a basic implementation of probability of extinction as 1 - exp(-rate * dt). Models 'II' and 'IV' use a more detailed functional relationship between rate and probability, following Raup (1984): (µ*((exp((lambda-µ)*dt))-1))/(lambda*((exp((lambda-µ)*dt)))-µ). Modeltype 'V' implements a temporally uninformed model, estimating the probabilities directly (or rather their logit).
spec/ext/samp	are formulas specifying the particular model. Use spec=~time for temporally variable rates, spec~div for diversity dependence. Other drivers should have names as entries into data
priorsMu	distribution used for prior for the baseline rates - a probability density function. Default is dnorm.
priorsCov	distribution used for prior for the driver coefficients - a probability density function. Default is dnorm.
priorsStd	distribution used for prior for log of the standard deviation of the random effects - a probability density function. Default is dunif.

Value

The function returns an object of class CMR_model. This most important output is out$probfun which is a function for the posterior. This output can be fed directly into the sampler MCMC_CMR. It also returns $Obs and $dts as well as the full call (exlcluding default settings), $call

Examples

sim <- sim_BD(ext = function(t,n){0.01+0.03*t},samp = 0.8,n_init = 10)
plot(sim)
mod1 <- make.BayesCMR(sim$FosRec>0)
fit <- MCMC_CMR(mod1)
plot(fit)
summary(fit)
matplot(fit$Chain[,1:3],type="l")

josteist/Compadre documentation built on June 3, 2021, 1:02 p.m.