get.parsamp {timedeppar}

Get a sample of lists of constant and time-dependent parameters from inference results of infer.timedeppar

Description

This function produces a sample of parameter sets for past and potentially future time points based on the results of class timedeppar generated by Bayesian inference with the function infer.timedeppar. For time points used for inference, the sample is a sub-samble of the Markov chain, for future time points of time-dependent parameters it is a random sample based on the corresponding Ornstein-Uhlenbeck parameters and constrained at there initial point to the end point of the sub-sample.

Usage

get.parsamp(x, samp.size = 1000, n.burnin = 0, times.new = numeric(0))

Arguments

x

results from the function infer.timedeppar of class timedeppar.

samp.size

size of the produced sample constructed from the Markov chain stored in the object of class timedeppar omitting the adaptation and burnin phases.

n.burnin

number of Markov chain points to omit for density and pairs plots (number of omitted points is max(control$n.adapt,n.burnin)).

times.new

vector of time points to predict for. If no time points are provided, sampling is only from the inference Markov chain; if time points are provided, they need to be increasing and start with a larger value than the time points used for inference. In the latter case, time-dependent parameters are sampled for the future points and appended to the inferred part of the time-dependend parameter.

Value

list of
param.maxpost: list of constant and time-dependent parameters corresponding to the maximum posterior solution for inference (no extrapolation to the future).
param.maxlikeli: list of constant and time-dependent parameters corresponding to the solution with maximum observation likelihood found so far.
param.list: list of length samp.size containing lists of constant and time-dependent parameters; for time-dependent parameters sub-sample of the Markov chain for past time points, sample from Ornstein-Uhlenbeck processes conditioned at the initial point for future time points (see argument times.new).
param.const: sub-sample of constant parameters.
param.timedep: list of sub-samples of time-dependent parameters.
param.ou: sub-sample of Ornstein-Uhlebeck parameters of the time-dependent parameter(s).
ind.timedeppar: indices of time-dependent parameters in the parameter lists.
ind.sample: indices of the stored, thinned sample defining the sub-sample.
ind.chain: indices of the original non-thinned Markov chain defining the sub-sample.
dot.args: ... arguments passed to infer.timedeppar; to be re-used for new model evaluations.