R Graphical Manual

Browse All

Last data update: 2014.03.03

R: Runs the Compound Poisson-Gamma chronology model of Haslett...

Bchronology

R Documentation

Runs the Compound Poisson-Gamma chronology model of Haslett and Parnell (2008)

Description

Fits a non-parametric chronology model to age/position data according to the Compound Poisson-Gamma model defined by Haslett and Parnell (2008). This version uses a slightly modified Markov chain Monte Carlo fitting algorithm which aims to converge quicker and requires fewer iterations. It also a slightly modified procedure for identifying outliers

Usage

Bchronology(ages, ageSds, positions, positionThicknesses = rep(0, length(ages)), 
calCurves = rep("intcal13", length(ages)), ids = NULL, 
outlierProbs = rep(0.01, length(ages)), predictPositions = seq(min(positions), 
max(positions), length = 100), pathToCalCurves = system.file('data', 
package = "Bchron"), iterations = 10000, burn = 2000, thin = 8, 
extractDate = 1950 - as.numeric(format(Sys.time(), "%Y")), maxExtrap = 500, 
thetaMhSd = 0.5, muMhSd = 0.1, psiMhSd = 0.1, ageScaleVal = 1000, 
positionScaleVal = 100)

Arguments

`ages`	A vector of ages (most likely 14C)
`ageSds`	A vector of 1-sigma values for the ages given above
`positions`	Position values (e.g. depths) for each age
`positionThicknesses`	(optional) Thickness values for each of the positions. By default set to zero
`calCurves`	A vector of values containing either 'intcal13', 'shcal13', 'marine13', or 'normal'. Should be the same length the number of ages supplied. Non-standard calibration curves can be used provided they are supplied in the same format as those previously mentioned and are placed in the same directory. Normal indicates a normally-distributed (non-14C) age.
`ids`	(optional) ID names for each age
`outlierProbs`	(optional) A vector of prior outlier probabilities, one for each age. Defaults to 0.01
`predictPositions`	(optional) A vector of positions (e.g. depths) at which predicted age values are required. Defaults to a sequence of length 100 from the top position to the bottom position
`pathToCalCurves`	(optional) File path to where the calibration curves are located. Defaults to the system directory where the 3 standard calibration curves are stored.
`iterations`	(optional) The number of iterations to run the procedure for
`burn`	(optional) The number of starting iterations to discard
`thin`	(optional) The step size for every iteration to keep beyond the burnin
`extractDate`	(optional) The top age of the core. Used for extrapolation purposes so that no extrapolated ages go beyond the top age of the core. Defaults to the current year
`maxExtrap`	(optional) The maximum number of extrapolations to perform before giving up. Useful for when large amounts of extrapolation are required, i.e. some of the predictPositions are a long way from the dated positions
`thetaMhSd`	(optional) The Metropolis-Hastings standard deviation for the age parameters
`muMhSd`	(optional) The Metropolis-Hastings standard deviation for the Compound Poisson-Gamma mean
`psiMhSd`	(optional) The Metropolis-Hastings standard deviation for the Compound Poisson-Gamma scale
`ageScaleVal`	(optional) A scale value for the ages. Bchronology works best when the ages are scaled to be approximately between 0 and 100. The default value is thus 1000 for ages given in years.
`positionScaleVal`	(optional) A scale value for the positions. Bchronology works best when the positions are scaled to be approximately between 0 and 100. The default value is thus 100 for positions given in cm.

Details

The Bchronology function fits a compound Poisson-Gamma distribution to the increments between the dated levels. This involves a stochastic linear interpolation step where the age gaps are Gamma distributed, and the position gaps are Exponential. Radiocarbon and non-radiocarbon dates (including outliers) are updated within the function also by MCMC.

Value

A list of class BchronologyRun which include elements:

`theta`	The posterior estimated values of the ages
`phi`	The posterior estimated outlier values (1=outlier, 2=not outlier). The means of this parameter give the posterior estimated outlier probabilities
`mu`	The posterior values of the Compound Poisson-Gamma mean
`psi`	The posterior values of the Compound Poisson-Gamma scale
`thetaPredict`	The posterior estimated ages for each of the values in predictPosition
`predictPositions`	The positions at which estimated ages were required
`calAges`	The calibrated ages as output from `BchronCalibrate`

Author(s)

Andrew Parnell <andrew.parnell@ucd.ie>

References

Haslett, J., and Parnell, A. C. (2008). A simple monotone process with application to radiocarbon-dated depth chronologies. Journal of the Royal Statistical Society, Series C, 57, 399-418.

Parnell, A. C., Haslett, J., Allen, J. R. M., Buck, C. E., and Huntley, B. (2008). A flexible approach to assessing synchroneity of past events using Bayesian reconstructions of sedimentation history. Quaternary Science Reviews, 27(19-20), 1872-1885.

Examples

## Not run: 
# Data from Glendalough
data(Glendalough)

# Run in Bchronology - all but first age uses intcal13
GlenOut = Bchronology(ages=Glendalough$ages,ageSds=Glendalough$ageSds,
calCurves=Glendalough$calCurves,positions=Glendalough$position,
positionThicknesses=Glendalough$thickness,ids=Glendalough$id,
predictPositions=seq(0,1500,by=10))

# Summarise it a few different ways
summary(GlenOut) # Default is for quantiles of ages at predictPosition values
summary(GlenOut, type='convergence') # Check model convergence
summary(GlenOut, type='outliers') # Look at outlier probabilities

# Predict for some new positions
predictAges = predict(GlenOut, newPositions = c(150,725,1500), newPositionThicknesses=c(5,0,20))

# Plot the output
plot(GlenOut,main="Glendalough",xlab='Age (cal years BP)',ylab='Depth (cm)',las=1)
## End(Not run)