metropolisDate: Estimate segment dates

In AustralianAntarcticDataCentre/datemoss: Chronologies and Growth Rates of Moss

Description

Estimate segment times and growth rates by MCMC

Usage

metropolisDate(ts.init, tmin, len, pmc, Year, PMC, iters = 2000L,
  thin = 25L, chains = if (is.list(ts.init)) length(ts.init) else 1L,
  log.prior = NULL, alpha = 0.01, beta = 0.01, verbose = interactive())

metropolisDate0(ts.init, tmin, len, pmc, Year, PMC, sigma = 4,
  iters = 2000L, thin = 25L, chains = if (is.list(ts.init))
  length(ts.init) else 1L, log.prior = NULL, verbose = interactive())

Arguments

ts.init	a vector or a list of vectors of initial date estimates, as generated by generateInitial
tmin	the minimum possible date
len	lengths of the segments (in timewise order, oldest first)
pmc	mean percentage modern carbon concentration of the segments (in timewise order, oldest first)
Year	calibration data - (fractional) year of atmospheric carbon measurement
PMC	calibration data - percent modern carbon
iters	number of samples to draw.
thin	rate at which to thin samples.
chains	number of chains to sample.
log.prior	function to calculate the contribution from each segment to the log prior for growth rates, given segment times and lengths.
alpha	shape parameter for the Gamma prior for tau
beta	rate parameter for the Gamma prior for tau
verbose	report progress at prompt?
sigma	standard deviation of PMC errors

Details

The two functions metropolisDate0 and metropolisDate estimate the segment cut times by Metropolis sampling. The metropolisDate0 variant assumes the standard deviation of the PMC errors (sigma) is fixed, while metropolisDate Gibbs samples for sigma.

The user must supply the lengths len and mean carbon concentrations pmc of the segments. These must be supplied in timewise order - oldest segment first and the segments must be contiguous, but the concentration data may contain missing values.

The user must also supply atmospheric radiocarbon calibration data as two vectors, Year the time of measurement in (fractional) years and PMC the recorded percentage of modern carbon.

The user may also supply a function that computes the contributions of each segment to the log prior. This function takes two arguments, ts a vector of the (n+1) segment endpoint times and len the lengths of the n segments. This function must compute the log of the (non-normalized) contributions to the prior from each segment.

Value

metropolisDate0 returns a list containing a coda object describing the segment times. metropolisDate returns a list containing a coda object describing the segment times and a coda object describing

AustralianAntarcticDataCentre/datemoss documentation built on May 5, 2019, 8:14 a.m.