This function generates a reference function from a mcmc chain for use in marginal likelihood estimation.
make.refFn(chain, model, priorFn, burnin = 0.3, plot = TRUE)An mcmc chain produced by bayou.mcmc() and loaded with load.bayou()
A string specifying the model ("OU", "QG", "OUrepar") or a model parameter list
The prior function used to generate the mcmc chain
The proportion of the mcmc chain to be discarded when generating the reference function
Logical indicating whether or not a plot should be created
Returns a reference function of class "refFn" that takes a parameter list and returns the log density
given the reference distribution. If plot=TRUE, a plot is produced showing the density of variable parameters
and the fitted distribution from the reference function (in red).
Distributions are fit to each mcmc chain and the best-fitting distribution is chosen as
the reference distribution for that parameter using the method of Fan et al. (2011). For positive
continuous parameters alpha, sigma^2, halflife, Vy, w2, Ne, Log-normal, exponential, gamma and weibull
distributions are fit. For continuous distributions theta, Normal, Cauchy and Logistic distributions
are fit. For discrete distributions, k, negative binomial, poisson and geometric distributions are fit.
Best-fitting distributions are determined by AIC.